U.S. patent number 11,170,473 [Application Number 17/080,543] was granted by the patent office on 2021-11-09 for method and apparatus for streaming data.
This patent grant is currently assigned to SAMSUNG ELECTRONICS CO., LTD.. The grantee listed for this patent is SAMSUNG ELECTRONICS CO., LTD.. Invention is credited to Kwangpyo Choi, Myungjin Eom, Youngo Park, Yumi Sohn.
United States Patent |
11,170,473 |
Park , et al. |
November 9, 2021 |
Method and apparatus for streaming data
Abstract
A terminal for receiving streaming data may receive information
of a plurality of different quality versions of an image content;
request, based on the information, a server for a version of the
image content from among the plurality of different quality
versions of the image content; when the requested version of the
image content and artificial intelligence (AI) data corresponding
to the requested version of the image content are received,
determines whether to perform AI upscaling on the received version
of the image content, based on the AI data; and based on a result
of the determining whether to perform AI upscaling, performs AI
upscaling on the received version of the image content through a
upscaling deep neural network (DNN) that is trained jointly with a
downscaling DNN of the server.
Inventors: |
Park; Youngo (Suwon-si,
KR), Sohn; Yumi (Suwon-si, KR), Eom;
Myungjin (Suwon-si, KR), Choi; Kwangpyo
(Suwon-si, KR) |
Applicant: |
Name |
City |
State |
Country |
Type |
SAMSUNG ELECTRONICS CO., LTD. |
Suwon-si |
N/A |
KR |
|
|
Assignee: |
SAMSUNG ELECTRONICS CO., LTD.
(Suwon-si, KR)
|
Family
ID: |
70281205 |
Appl.
No.: |
17/080,543 |
Filed: |
October 26, 2020 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20210056666 A1 |
Feb 25, 2021 |
|
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
16659061 |
Oct 21, 2019 |
10817986 |
|
|
|
Foreign Application Priority Data
|
|
|
|
|
Oct 19, 2018 [KR] |
|
|
10-2018-0125406 |
Apr 8, 2019 [KR] |
|
|
10-2019-0041110 |
Jun 27, 2019 [KR] |
|
|
10-2019-0077250 |
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
H04N
19/46 (20141101); H04N 21/440263 (20130101); H04N
19/463 (20141101); H04N 19/85 (20141101); H04L
65/607 (20130101); H04N 19/132 (20141101); H04N
21/23439 (20130101); H04N 21/8193 (20130101); H04N
19/172 (20141101); H04N 19/59 (20141101); H04N
21/251 (20130101); H04N 19/117 (20141101); H04N
19/80 (20141101); G06T 3/4046 (20130101); H04L
65/602 (20130101); H04N 21/26258 (20130101); H04N
19/154 (20141101); H04L 65/4069 (20130101); H04L
65/80 (20130101); H04N 21/234363 (20130101); H04N
19/14 (20141101); H04N 19/146 (20141101); H04N
19/166 (20141101) |
Current International
Class: |
G06T
3/40 (20060101); H04N 19/85 (20140101); H04L
29/06 (20060101) |
Field of
Search: |
;382/156-158,232-233,298-300 ;348/14.13 |
References Cited
[Referenced By]
U.S. Patent Documents
Foreign Patent Documents
|
|
|
|
|
|
|
2008-527810 |
|
Jul 2008 |
|
JP |
|
2009-540625 |
|
Nov 2009 |
|
JP |
|
2012-191250 |
|
Oct 2012 |
|
JP |
|
10-0224801 |
|
Oct 1999 |
|
KR |
|
10-0286443 |
|
Apr 2001 |
|
KR |
|
10-1425602 |
|
Jul 2014 |
|
KR |
|
10-2014-0145560 |
|
Dec 2014 |
|
KR |
|
1020150135637 |
|
Dec 2015 |
|
KR |
|
10-2016-0036662 |
|
Apr 2016 |
|
KR |
|
10-2016-0080929 |
|
Jul 2016 |
|
KR |
|
10-2017-0059040 |
|
May 2017 |
|
KR |
|
10-2017-0100045 |
|
Sep 2017 |
|
KR |
|
10-2018-0001428 |
|
Jan 2018 |
|
KR |
|
10-2018-0043154 |
|
Apr 2018 |
|
KR |
|
10-2018-0052651 |
|
May 2018 |
|
KR |
|
10-1885855 |
|
Aug 2018 |
|
KR |
|
10-2018-0100976 |
|
Sep 2018 |
|
KR |
|
10-2018-0108288 |
|
Oct 2018 |
|
KR |
|
2016205733 |
|
Dec 2016 |
|
WO |
|
2017-036370 |
|
Mar 2017 |
|
WO |
|
2018/140294 |
|
Aug 2018 |
|
WO |
|
2018/143992 |
|
Aug 2018 |
|
WO |
|
2019/220095 |
|
Nov 2019 |
|
WO |
|
Other References
Li, Yue, et al. "Learning a convolutional neural network for image
compact-resolution." IEEE Transactions on Image Processing 28.3
(2018): 1092-1107. (Year: 2018). cited by examiner .
Sani, Yusuf, Andreas Mauthe, and Christopher Edwards. "Adaptive
bitrate selection: A survey." IEEE Communications Surveys &
Tutorials 19.4 (2017): 2985-3014. (Year: 2017). cited by examiner
.
Aklson, Alex. "Shallow Versus Deep Neural Networks." Introduction
to Deep Learning & Neural Networks with Keras, Coursera.
https://www.coursera.org/lecture/introduction-to-deep-learning-with-keras-
/shallow-versus-deep-neural-networks-3pKHn. Accessed Nov. 26, 2019.
(Year: 2019). cited by applicant .
Communication (PCT/ISA/220, PCT/ISA/210 & PCT/ISA/237) dated
Dec. 4, 2019 by the International Searching Authority in
counterpart International Patent Application No. PCT/KR2019/010645.
cited by applicant .
Communication (PCT/ISA/220, PCT/ISA/210 & PCT/ISA/237) dated
Feb. 21, 2020 by the International Searching Authority in
counterpart International Patent Application No. PCT/KR2019/013733.
cited by applicant .
Communication (PCT/ISA/220, PCT/ISA/210 & PCT/ISA/237) dated
Feb. 5, 2020 by the International Searching Authority in
counterpart International Patent Application No. PCT/KR2019/013483.
cited by applicant .
Communication (PCT/ISA/220, PCT/ISA/210 & PCT/ISA/237) dated
Jan. 23, 2020 by the International Searching Authority in
counterpart International Patent Application No. PCT/KR2019/012836.
cited by applicant .
Communication (PCT/ISA/220, PCT/ISA/210 & PCT/ISA/237) dated
Jan. 29, 2020 by the International Searching Authority in
counterpart International Patent Application No. PCT/KR2019/013595.
cited by applicant .
Communication (PCT/ISA/220, PCT/ISA/210 & PCT/ISA/237) dated
Jan. 30, 2020 by the International Searching Authority in
counterpart International Patent Application No. PCT/KR2019/013421.
cited by applicant .
Giannaopoulos, Michalis et al., "Convolutional Neural Networks for
Video Quality Assessment", arXiv:1809.10117v1, [eess.IV], Sep. 26,
2018, pp. 1-12. (14 pages total). cited by applicant .
Gorodilov, Artem, Dmitriy Gavrilov, and Dmitriy Schelkunov. "Neural
Networks for Image and Video Compression." 2018 International
Conference on Artificial Intelligence Applications and Innovations
(IC-AIAI). IEEE. (Year: 2018). cited by applicant .
Jeong, Woojin et al., "Efficient Super-Resolution Method for Single
Image based on Deep Neural Networks", Journal of The Institute of
Electronics and Information Engineers, vol. 55, No. 6, pp. 779-786,
Jun. 2018. (10 pages total). cited by applicant .
Jiang, Feng, et al. "An end-to-end compression framework based on
convolutional neural networks." IEEE Transactions on Circuits and
Systems for Video Technology 28.10 (2017): 3007-3018. (Year: 2017).
cited by applicant .
Ki, Sehwan et al., "A Study on the Convolution Neural Network based
on Blind High Dynamic Range Image Quality Assessment", Proceedings
of Symposium of the Korean Institute of Communications and
Information Sciences, 2018, pp. 1529-1530. (4 pages total). cited
by applicant .
Le Callet, Patrick et al., "A Convolutional Neural Network Approach
for Objective Video Quality Assessment", IEEE Transactions on
Neural Networks, vol. 17, No. 5, Sep. 2006, pp. 1316-1327. (14
pages total). cited by applicant .
Matthieu Courbariaux et al., "BinaryConnect: Training Deep Neural
Networks with binary weights during propagations", Apr. 18, 2016,
pp. 1-9 (9 pages total). cited by applicant .
Mohammad Rastegari et al., "XNOR-Net: ImageNet Classification Using
Binary Convolutional Neural Networks", Sep. 17, 2016, Computer
Vision--ECCV 2016. ECCV 2016. Lecture Notes in Computer Science,
vol. 9908, (17 pages total). cited by applicant .
PCT/ISA/210 & PCT/ISA/237 dated Aug. 6, 2019 by the
International Searching Authority in counterpart International
Patent Application No. PCT/KR2019/004171. cited by applicant .
PCT/ISA/210 & PCT/ISA/237 dated Jan. 22, 2020 by the
International Searching Authority in counterpart International
Patent Application No. PCT/KR2019/013344. cited by applicant .
PCT/ISA/210 & PCT/ISA/237 dated Jan. 7, 2020 by the
International Searching Authority in counterpart International
Patent Application No. PCT/KR2019/012510. cited by applicant .
Sergey Ioffe et al., "Batch Normalization: Accelerating Deep
Network Training by Reducing Internal Covariate Shift", Proceedings
of the 32nd International Conference on Machine Learning 2015, vol.
37, pp. 448-456, 2015. (11 pages total). cited by applicant .
Seungbin Lee et al., "Performance Analysis of Convolution Neural
Network and Generative Adversarial Network for Super Resolution",
Journal of the Korean Institute of Information Scientists and
Engineers, Jun. 2017, pp. 931-933. (6 pages total). cited by
applicant .
Shuai, Hong-Han, et al. "Mobiup: An upsampling-based system
architecture for high-quality video streaming on mobile devices."
IEEE Transactions on Multimedia 13.5 (2011): 1077-1091. (Year:
2011). cited by applicant .
Yeo, Hyunho, et al. "Neural adaptive content-aware internet video
delivery." 13th [USENIX} Symposium on Operating Systems Design and
Implementation {{OSDI} 18).2018. (Year: 2018). cited by applicant
.
Yeo, Hyunho, sunghyun Do, and Dongsu Han. "How will Deep Learning
Change Internet Video Delivery?." Proceedings of the 16th ACM
Workshop on Hot Topics in Networks. ACM, 2017. (Year: 2017). cited
by applicant .
Stockhammer, Thomas. "Dynamic adaptive streaming over
HTTP--standards and design principles." Proceedings of the second
annual ACM conference on Multimedia systems. 2011. (Year: 2011).
cited by applicant .
Afonso, Mariana, Fan Zhang, and David R. Bull. "Spatial resolution
adaptation framework for video compression." Applications of
Digital Image Processing XLI. vol. 10752. International Society for
Optics and Photonics, 2018 (Year: 2018). cited by applicant .
Kim, Jaehwan, et al., "Dynamic frame resizing with convolutional
neural network for efficient video compression." Applications of
Digital Image Processing XL. vol. 10396. International Society for
Optics and Photonics, 2017. (Year: 2017). cited by applicant .
Adhikari, Vijay Kumar, et al. "Unreeling netflix: Understanding and
improving multi-cdn movie delivery." 2012 Proceedings IEEE INFOCOM.
IEEE, 2012. (Year: 2012). cited by applicant .
Bing, Benny. Next-Generation Video Coding and Streaming, John Wiley
& Sons, Incorporated, 2015, pp. 245-310. ProQuest Ebook
Central,
http://ebookcentral.proquest.com/lib/uspto-ebooks/detail.action?docID=440-
6551. (Year: 2015). cited by applicant .
Cho, et al., "A Technical Analysis of Deep Learning based Image and
Video Compression", 2018, Journal of Broadcast Engineering Society,
et al., vol. 23, Issue 3, 12 pages total. cited by applicant .
Johnston, et al., "Improved Lossy Image Compression with Priming
and Spatially Adaptive Bit Rates for Recurrent Networks", 2017, 9
pages total. cited by applicant .
Louizos et al., "Relaxed Quantization for Discretized Neural
Networks", 2018, 14 pages total. cited by applicant .
Kim, et al., "Deeply-Recursive Convolutional Network for Image
Super-Resolution", 2016, 9 pages total. cited by applicant .
Qiu, et al., "FReLU: Flexible Rectified Linear Units for Improving
Convolutional Neural Networks", 2018 6 pages total. cited by
applicant .
Mao, et al., "Image Restoration Using Convolutional Auto-encoders
with Symmetric Skip Connections" 2016, version 3, Computer Science,
22 pages total. cited by applicant .
Yahia, et al., "Frame Interpolation using Convolutional Neural
Networks on 2D animation", 2016, Universiteit van Amsterdam, 20
pages total. cited by applicant .
Skoneczny, et al., "Classical and neural methods of image sequence
interpolation". SPIE, 2000, 15 pages total. cited by applicant
.
Communication dated Dec. 22, 2020, issued by the Korean
Intellectual Property Office in Korean Application No.
10-2019-0066057. cited by applicant .
Communication dated Jan. 12, 2021, issued by the Korean
Intellectual Property Office in Korean Application No.
10-2019-0078343. cited by applicant .
Communication dated Jan. 13, 2021, issued by the Korean
Intellectual Property Office in Korean Application No.
10-2019-0078344. cited by applicant .
Communication dated Jan. 14, 2021, issued by the Korean
Intellectual Property Office in Korean Application No.
10-2020-7000378. cited by applicant .
Communication dated Dec. 10, 2020, issued by the Korean
Intellectual Property Office in Korean Application No.
10-2019-0041102. cited by applicant .
Communication dated Dec. 11, 2020, from the European Patent Office
in European Application No. 19 183 429.0. cited by applicant .
Communication dated Dec. 11, 2020, issued by the Korean
Intellectual Property Office in Korean Application No.
10-2019-0041109. cited by applicant .
Communication dated Dec. 16, 2020, issued by the Korean
Intellectual Property Office in Korean Application No.
10-2019-0062583. cited by applicant .
Communication dated Jan. 5, 2021, issued by the Korean Intellectual
Property Office in Korean Application No. 10-2019-0076569. cited by
applicant .
Communication dated Jan. 6, 2021, issued by the Korean Intellectual
Property Office in Korean Application No. 10-2019-0077250. cited by
applicant .
Communication dated Jun. 3, 2021, issued by the Korean Intellectual
Property Office in Korean Application No. 10-2019-0041102. cited by
applicant .
Communication dated Jun. 7, 2021, from the Intellectual Property
Office of India in Application No. 201924041670. cited by applicant
.
Li, Y., et al., "Convolutional Neural Network-Based Block
Up-Sampling for Intra Frame Coding", IEEE Transactions on Circuits
and Systems for Video Technology, vol. 28, No. 9, Sep. 2018, pp.
2316-2330. cited by applicant .
Communication dated Jun. 29, 2021 issued by the European Patent
Office in application No. 19872393.4. cited by applicant .
Communication dated Jun. 29, 2021 issued by the European Patent
Office in application No. 19873762.9. cited by applicant .
Communication dated Jul. 5, 2021 issued by the European Patent
Office in application No. 19873269.5. cited by applicant .
Communication dated Jun. 21, 2021 issued by the Korean Intellectual
Property Office in application No. 10-2019-0041109. cited by
applicant .
Communication dated Jun. 21, 2021 issued by the Korean Intellectual
Property Office in application No. 10-2019-0062583. cited by
applicant .
Communication dated Jun. 21, 2021 issued by the Korean Intellectual
Property Office in application No. 10-2019-0066057. cited by
applicant .
Lucas Theis, et al., "Lossy Image Compression with Compressive
Autoencoders", ARXIV.org., Cornell University Library, XP080753545,
Mar. 1, 2017. cited by applicant .
Anonymous et al., "Machine Learning: Why Do We Need Floats For
Using Neural Networks?--Artificial Intelligence Stack Exchange",
XP055822829, Aug. 16, 2018, pp. 1-8. cited by applicant .
Suyog Gupta, et al., Deep Learning With Limited Numerical
Precision, ARXIV.org, Cornell University, XP080677454, Feb. 9,
2015; Abstract, sections 3-3.2, 4.1.2 and 4.2. cited by applicant
.
Supplementary European Search Report dated Jul. 20, 2021 issued in
application No. 19874036.7-1230. cited by applicant .
Office Action issued in Korean Patent Application No.
10-2019-0077250, dated Jul. 21, 2021. cited by applicant .
Office Action issued in Korean Patent Application No.
10-2019-0076569, dated Jul. 21, 2021. cited by applicant .
Foreign Action issued in Korean Patent Application No. 2019-0062583
dated Aug. 9, 2021. cited by applicant .
Foreign Action issued in Korean Patent Application No. 2019-0066057
dated Aug. 9, 2021. cited by applicant .
Office Action issued in Korean Patent Application No.
10-2019-0076569, dated Sep. 16, 2021. cited by applicant .
Office Action issued in Korean Patent Application No.
10-2019-0077250, dated Sep. 16, 2021. cited by applicant .
Extended European Search Report issued in European Patent
Application No. 19873223.2 dated Oct. 4, 2021. cited by applicant
.
Sajjadi, Medhi, S., "Enhance Net: Single Image Super-Resolution
Through Automated Texture Synthesis", 2017 IEEE International
Conference on Computer Vision, 2017, pp. 4501-4510. cited by
applicant .
Guo, Tiantong, et al., 37 Deep Learning Based Image
Super-Resolution with Coupled Backpropagation, Department of
Electrical Engineering, The Pennsylvania State University, 2016,
237-241. cited by applicant.
|
Primary Examiner: Goradia; Shefali D
Attorney, Agent or Firm: Sughrue Mion, PLLC
Parent Case Text
CROSS-REFERENCE TO RELATED APPLICATIONS
This application is a continuation of U.S. patent application Ser.
No. 16/659,061, filed on Oct. 21, 2019, and is based on and claims
priority under 35 U.S.C. .sctn. 119 to Korean Patent Application
Nos. 10-2018-0125406, filed on Oct. 19, 2018, and 10-2019-0041110,
filed on Apr. 8, 2019 and 10-2019-0077250, filed on Jun. 27, 2019,
in the Korean Intellectual Property Office, the disclosures of
which are incorporated herein by reference in their entireties.
Claims
What is claimed is:
1. A server for streaming data, the server comprising: at least one
processor, when executing one or more instructions stored in the
server, configured to: receive, from an electronic device, a first
request for image data of a first quality, in response to the first
request, control the transceiver to transmit, to the electronic
device, first artificial intelligence (AI) data and the image data
of the first quality that has been AI encoded through a downscaling
neural network (NN) of the server, the first AI data related to AI
downscaling an original image to a first image through the
downscaling NN based on NN setting information for downscaling, the
first NN setting information for downscaling being selected from a
plurality of NN setting information for downscaling; identify a
state of a network between the electronic device and the server;
and transmit, to the electronic device, second AI data and the
image data of a second quality based on the state of the network,
wherein a first AI upscaling on the image data of the first quality
is performed by the electronic device through a first upscaling NN
that is set with first NN setting information for upscaling at the
electronic device, the first NN setting information for upscaling
corresponding to the first AI data, and a second AI upscaling on
the image data of the second quality is performed by the electronic
device through a second upscaling NN that is set with second NN
setting information for upscaling, wherein the first NN setting
information for upscaling is selected based on the first AI data
from a plurality of NN setting information for upscaling, and the
second NN setting information for upscaling is selected based on
the second AI data from the plurality of NN setting information for
upscaling, wherein the first AI data comprises an index indicating
the first NN setting information for downscaling among the
plurality of NN setting information for downscaling.
2. The server of claim 1, wherein the at least one processor is
further configured to execute the one or more instructions to
receive a second request for the image data of the second quality
that is determined based on either one or both of AI scale
conversion information and quality information of each of a
plurality of image data of different qualities.
3. The server of claim 1, wherein the image data of the second
quality corresponds to the state of the network and is determined
based on capability information comprising information indicating
whether AI upscaling technology is supported by the electronic
device and information about an AI upscale level supported by the
electronic device.
4. The server of claim 1, wherein the at least one processor is
further configured to execute the one or more instructions to
provide the electronic device with an identifier of the server.
5. The server of claim 1, wherein the at least one processor is
further configured to execute the one or more instructions to
receive, from the electronic device, a second request for the image
data of the second quality from among a plurality of image data of
different qualities, based on a user input selecting at least one
of a second bitrate or a second resolution related to the second
quality.
6. The server of claim 1, wherein the first AI data further
comprises information related to at least one of a resolution
difference between the original image and the first image, a
bitrate regarding the image data of the first quality, a
quantization parameter regarding the image data of the first
quality, a resolution of the first image, or a codec type used to
encode the first image.
7. The server of claim 1, wherein the first NN setting information
for downscaling comprises a parameter set for downscaling.
8. The server of claim 7, wherein the parameter set for downscaling
comprises at least one of a weight and a bias of a NN.
9. An electronic device for processing streaming data, the
electronic device comprising: at least one processor, when
executing one or more instructions stored in the electronic device,
configured to: request a server to transmit image data of a first
quality; receive the image data of the first quality and first
artificial intelligence (AI) data as a response to the requesting
the server to transmit the image data of the first quality, the
first AI data related to AI downscaling an original image to a
first image through a downscaling neural network (NN) based on
first NN setting information for downscaling, the first NN setting
information for downscaling being selected from a plurality of NN
setting information for downscaling; select, based on the first AI
data, first NN setting information for upscaling from a plurality
of NN setting information for upscaling, the first NN setting
information for upscaling corresponding to the received first AI
data; perform a first AI upscaling on the received image data of
the first quality, through an upscaling NN of the electronic device
that is set with the selected first NN setting information for
upscaling; receive image data of a second quality and second AI
data based on a result of identifying of a state of a network by
the server; select, based on the second AI data, second NN setting
information for upscaling from the plurality of NN setting
information for upscaling, the second NN setting information for
upscaling corresponding to the received second AI data; and perform
a second AI upscaling on the received image data of the second
quality, through the upscaling NN that is set with the selected
second NN setting information for upscaling, and wherein the first
AI data comprises an index indicating the first NN setting
information for downscaling among the plurality of NN setting
information for downscaling.
10. The electronic device of claim 9, wherein the at least one
processor is further configured to execute the one or more
instructions to: determine, based on the first AI data, whether AI
downscaling has been performed on the received image data of the
first quality through the downscaling NN of the server, and based
on the AI downscaling having been performed on the received image
data of the first AI data, perform the first AI upscaling on the
received image data of the first quality.
11. The electronic device of claim 9, wherein the at least one
processor is further configured to execute the one or more
instructions to receive information of a plurality of image data of
different qualities, and when the at least one processor requests
the server to execute the one or more instructions to transmit the
image data of the first quality, the at least one processor is
configured to execute the one or more instructions to request,
based on at least the information, the server to transmit the image
data of the first quality, from among the plurality of image data
of different qualities.
12. The electronic device of claim 11, wherein the information of
the plurality of image data includes quality information and AI
scale conversion information of the plurality of image data, and
wherein the at least one processor is further configured to execute
the one or more instructions to request the image data of the first
quality, based on either one or both of the quality information and
the AI scale conversion information.
13. The electronic device of claim 9, wherein when the at least one
processor is configured to execute the one or more instructions to
receive the image data of the second quality and second AI data,
the at least one processor is configured to execute the one or more
instructions to receive the image data of the second quality and
second AI data based on the result of the identifying of the state
of the network and a result of identifying of capability
information comprising information indicating whether the second AI
upscaling is supported by the electronic device and information
about an AI upscale level supported by the electronic device.
14. The electronic device of claim 9, wherein the server is a
content provider server, and wherein the at least one processor is
further configured to execute the one or more instructions to
request a service server to provide the electronic device with
information of a plurality of image data, and to receive, from the
service server an identifier of the content provider server and the
information of the plurality of image data.
15. The electronic device of claim 9, wherein the image data of the
second quality comprises image data corresponding to the state of
the network based on at least one of a Bit Rate related to the
image data and a Bit Error Rate (BER) related to the image
data.
16. The electronic device of claim 9, wherein the state of the
network between the electronic device and the server is based on at
least one of a delay, a bit rate, a bit error rate, a packet loss,
a time out, or capability of the electronic device.
17. The electronic device of claim 9, wherein the at least one
processor is configured to execute the one or more instructions to
request the server to transmit the image data of the second quality
based on a user input selecting at least one of a second bitrate or
a second resolution related to the second quality, and the at least
one processor is configured to execute the one or more instructions
to receive the image data of the second quality and second AI data
as a response to the requesting the server to transmit the image
data of the second quality based on the result of identifying of
the state of the network by the server.
18. The electronic device of claim 9, wherein the first NN setting
information for downscaling comprises a parameter set for
downscaling.
19. The electronic device of claim 18, wherein the parameter set
for downscaling comprises at least one of a weight and a bias of a
NN.
20. A non-transitory computer-readable recording medium having
recorded thereon instructions, which when executed by an electronic
device, cause the electronic device to perform first operations
comprising: requesting a server to transmit image data of a first
quality; receiving the image data of the first quality and first
artificial intelligence (AI) data as a response to the requesting
the server to transmit the image data of the first quality, the
first AI data related to AI downscaling an original image to a
first image through a downscaling neural network (NN) based on
first NN setting information for downscaling, the first NN setting
information for downscaling being selected from a plurality of NN
setting information for downscaling; selecting, based on the first
AI data, first NN setting information for upscaling from a
plurality of NN setting information for upscaling, the first NN
setting information for upscaling corresponding to the received
first AI data; performing a first AI upscaling on the received
image data of the first quality, through an upscaling NN of the
electronic device that is paired with the downscaling NN of the
server and that is set with the selected first NN setting
information for upscaling; receiving image data of a second quality
and second AI data based on a result of identifying of a state of a
network by the server; selecting, based on the second AI data,
second NN setting information for upscaling from the plurality of
NN setting information for upscaling, the second NN setting
information for upscaling corresponding to the received second AI
data; performing a second AI upscaling on the received image data
of the second quality, through the upscaling NN that is paired with
the downscaling NN of the server and that is set with the selected
second NN setting information for upscaling, and wherein the first
AI data comprises an index indicating the first NN setting
information for downscaling among the plurality of NN setting
information for downscaling.
Description
BACKGROUND
1. Field
The disclosure relates to data streaming technology. More
particularly, the disclosure relates to a method and apparatus for
adaptively streaming image data artificial intelligence
(AI)-encoded by using a deep neural network (DNN).
2. Description of Related Art
A scheme for transmitting image data through a network includes a
download scheme and a streaming scheme. The streaming scheme refers
to a scheme for transmitting, by a server, image data in real time,
and reproducing, by a terminal, received image data in real
time.
Unlike the download scheme in which reproduction of image data is
started after the image data is completely transceived, i.e.,
completely transmitted and received, according to the streaming
scheme, image data is transceived and reproduced in real time via a
logic channel established between a server and a terminal, and thus
a Quality of Service (QoS) of image data reproduction may be
maintained while reflecting a change in a streaming
environment.
Artificial intelligence (AI) systems are computer systems for
implementing human-level intelligence. Unlike general rule-based
smart systems, the AI systems autonomously learn and make
decisions, and thus improve their capabilities. The more the AI
systems are used, the more recognition rates of the AI systems
increase and the more accurately the AI systems understand user
preferences. As such, the general rule-based smart systems may be
replaced by deep-learning-based AI systems.
As interest in the AI systems increases, many attempts are actively
being made to apply the AI systems to various technology fields.
For example, research is being conducted to converge the AI systems
with technology fields including image processing, data processing,
and the like.
SUMMARY
Provided are a method and apparatus for streaming data that is
artificial intelligence (AI)-encoded by using a deep neural network
(DNN).
Additional aspects will be set forth in part in the description
which follows and, in part, will be apparent from the description,
or may be learned by practice of the presented embodiments of the
disclosure.
According to embodiment of the disclosure, there is provided a
method of streaming data, including: receiving information of a
plurality of different quality versions of an image content;
requesting, based on the information, a server to transmit a first
version of the image content to a terminal, from among the
plurality of different quality versions of the image content;
receiving the first version of the image content and artificial
intelligence (AI) data corresponding to the first version of the
image content; determining whether to perform AI upscaling on the
first version of the image content, based on the AI data; based on
a result of the determining whether to perform the AI upscaling,
performing AI upscaling on the first version of the image content
through an upscaling deep neural network (DNN) that is trained
jointly with a downscaling DNN of the server; confirming a state of
a network between the terminal and the server; and requesting the
server to transmit a second version of the image content to the
terminal, from among the plurality of different versions of the
image content, according to the information of the plurality of
different quality versions of the image content and the state of
the network.
The method may further include determining, based on the AI data,
whether AI downscaling has been performed on the first version of
the image content through the downscaling DNN of the server, and
wherein, when it is confirmed that the AI downscaling has been
performed on the first version of the image content, the
determining of whether the AI downscaling has been performed may
include determining to perform the AI upscaling on the first
version of the image content.
The information of the plurality of different quality versions of
the image content may include quality information and AI scale
conversion information of the plurality of different quality
versions of the image content, and the requesting the server to
transmit the second version of the image content may include
requesting the second version of the image content corresponding to
the state of the network, based on either one or both of the
quality information, and the AI scale conversion information.
The method may further include determining the second version of
the image content corresponding to the state of the network, based
on capability information including information indicating whether
the AI upscaling is supported by the terminal and information about
an AI upscale level supported by the terminal.
The server may be a content provider server, wherein the method may
further include requesting a service server for the information of
the plurality of different quality versions of the image content,
and wherein the receiving the information of the plurality of
different quality versions of the image content may include
receiving, from the service server, the information of the
plurality of different quality versions of the image content and an
identifier of the content provider server.
According to embodiment of the disclosure, there is provided a
method of streaming data, including: receiving, from a terminal, a
request for a first version of an image content from among a
plurality of different quality versions of the image content of a
server; in response to the request, transmitting, to the terminal,
artificial intelligence (AI) data and the first version of the
image content that has been AI encoded through a downscaling DNN of
the server that is trained jointly with a upscaling DNN of the
terminal; and receiving, from the terminal, a request for a second
version of the image content from among the plurality of different
versions of the image content, according to a state of a network
between the terminal and the server.
The AI data may include information about the downscaling DNN that
has been applied to the AI-encoded image data.
The receiving the request for the second version of the image
content may include receiving the request for the second version of
the image content that may correspond to the state of the network
and may be determined based on either one or both of AI scale
conversion information and quality information of each of the
plurality of different versions of the image content.
The second version of the image content corresponds to the state of
the network and is determined from among the plurality of different
versions of the image content, based on capability information
comprising information indicating whether AI upscaling is supported
by the terminal and information about an AI upscale level supported
by the terminal.
The method may further include providing the terminal with an
identifier of the server.
According to embodiment of the disclosure, a terminal for receiving
streaming data, including: a memory storing one or more
instructions; and at least one processor configured to execute the
one or more instructions to: receive information of a plurality of
different quality versions of an image content; request, based on
the information, a server to transmit a first version of the image
content, from among the plurality of different versions of the
image content; receive the first version of the image content and
artificial intelligence (AI) data corresponding to the first
version of the image content; determine whether to perform AI
upscaling on the first version of the image content, based on the
AI data; based on a result of the determining whether to perform
the AI upscaling, perform the AI upscaling on the first version of
the image content through an upscaling deep neural network (DNN) of
the terminal that is trained jointly with a downscaling DNN of the
server; confirm a state of a network between the terminal and the
server; and request, based on the information, the server to
transmit a second version of the image content to the terminal,
from among the plurality of different versions of the image
content, according to the information of the plurality of different
quality versions of the image content and the state of the
network.
The at least one processor may be further configured to: determine,
based on the AI data, whether AI downscaling has been performed on
the first version of the image content through the downscaling DNN
of the server; and when it is confirmed that the AI downscaling has
been performed on the first version of the image content, determine
to perform the AI upscaling on the first version of the image
content.
The information of the plurality of different quality versions of
the image content may include quality information and AI scale
conversion information of the plurality of different quality
versions of the image content, and wherein the at least one
processor may be further configured to execute the one or more
instructions to request the second version of the image content
corresponding to the state of the network, based on either one or
both of the quality information and the AI scale conversion
information.
The at least one processor may be further configured to: determine
the second version of the image content corresponding to the state
of the network, based on capability information comprising
information indicating whether AI upscaling is supported by the
terminal and information about an AI upscale level supported by the
terminal.
The server may be a content provider server, and the at least one
processor may be further configured to execute the one or more
instructions to request a service server to provide the terminal
with the information of the plurality of different quality versions
of the image content, and to receive, from the service server an
identifier of the content provider server and the information of
the plurality of different quality versions of the image
content.
According to embodiment of the disclosure, there is provided server
for streaming data, including: a memory storing one or more
instructions; and at least one processor configured to execute the
one or more instructions to: receive, from a terminal, a request
for a first version of an image content, from among a plurality of
different quality versions of the image content of a server; in
response to the request, transmit, to the terminal, artificial
intelligence (AI) data and the first version of the image content
that has been AI encoded through a downscaling deep neural network
(DNN) that is trained jointly with an upscaling DNN of the
terminal; and receive, from the terminal, a request for a second
version of the image content from the plurality of different
versions of the image content, according to a state of a network
between the terminal and the server.
The AI data may include information about the downscaling DNN that
has been applied to the AI-encoded image data.
The at least one processor may be further configured to execute the
one or more instructions to receive the request for the second
version of the image content that corresponds to the state of the
network and is determined based on either one or both of AI scale
conversion information and quality information of each of the
plurality of different versions of the image content.
The second version of the image content may correspond to the state
of the network and may be determined based on capability
information comprising information indicating whether AI upscaling
is supported by the terminal and information about an AI upscale
level supported by the terminal.
The at least one processor may be further configured to execute the
one or more instructions to provide the terminal with an identifier
of the server.
According to embodiment of the disclosure, there is provided a
non-transitory computer-readable recording medium having recorded
thereon a program for executing the method of steaming data.
According to embodiment of the disclosure, there is provided a
terminal for streaming data, including: a memory storing one or
more instructions; and at least one processor configured to execute
the one or more instructions to: receive, from a server,
information of a plurality of different quality versions of an
image content; determine a state of a network between the terminal
and the server; determine a version of the image content, from
among the plurality of different quality versions of the image
content, based on the information of the plurality of different
quality versions of the image content and the state of the network,
and request the server to transmit the version of the image content
to the terminal; receive, from the server, the version of the image
content and artificial intelligence (AI) data indicating whether
the version of the image content is downscaled through a
downscaling deep neural network (DNN) of the server; and process
the version of the image content based on the AI data.
According to embodiment of the disclosure, there is provided a
server for streaming data, including: a memory storing one or more
instructions; and at least one processor configured to execute the
one or more instructions to: provide a terminal with information of
a plurality of different quality versions of a image content;
receive, from the terminal, a request for a version of the image
content, from among the plurality of different quality versions of
the image content, according to a state of a network between the
terminal and the server; and provide the terminal with the
requested version of the image content and artificial intelligence
(AI) data indicating whether the requested version of the image
content is downscaled through a downscaling deep neural network
(DNN) of the server.
BRIEF DESCRIPTION OF THE DRAWINGS
The above and other aspects, features, and advantages of certain
embodiments of the disclosure will be more apparent from the
following description taken in conjunction with the accompanying
drawings, in which:
A brief description of each drawing is provided to more fully
understand the drawing recited in the present specification.
FIG. 1 is a diagram for describing an artificial intelligence (AI)
encoding process and an AI decoding process, according to
embodiments;
FIG. 2 is a block diagram of a configuration of an AI decoding
apparatus according to embodiments;
FIG. 3 is a diagram showing a second deep neural network (DNN) for
performing AI up-scaling on a second image;
FIG. 4 is a diagram for describing a convolution operation by a
convolution layer;
FIG. 5 is a table showing a mapping relationship between several
pieces of image-related information and several pieces of DNN
setting information;
FIG. 6 is a diagram showing a second image including a plurality of
frames;
FIG. 7 is a block diagram of a configuration of an AI encoding
apparatus according to embodiments;
FIG. 8 is a diagram showing a first DNN for performing AI
down-scaling on an original image;
FIG. 9 is a diagram for describing a method of training a first DNN
and a second DNN;
FIG. 10 is a diagram for describing a training process of a first
DNN and a second DNN by a training apparatus;
FIG. 11 is a diagram of an apparatus for performing AI down-scaling
on an original image and an apparatus for performing AI up-scaling
on a second image;
FIG. 12 is a diagram for describing a concept of a streaming
system, according to embodiments of the disclosure;
FIG. 13A is a flowchart for describing a method of streaming data,
the method being performed by a server, according to embodiments of
the disclosure;
FIG. 13B is a flowchart for describing a method of streaming data,
the method being performed by a terminal, according to embodiments
of the disclosure;
FIG. 14A is a flowchart for describing a method of streaming data,
the method being performed by a server, according to embodiments of
the disclosure;
FIG. 14B is a flowchart for describing a method of streaming data,
the method being performed by the terminal, according to
embodiments of the disclosure;
FIG. 15A is a flowchart for describing a method of streaming data,
the method being performed by a server, according to embodiments of
the disclosure;
FIG. 15B is a flowchart for describing a method of streaming data,
the method being performed by the terminal, according to
embodiments of the disclosure;
FIG. 16 is a diagram for describing a method of performing
streaming between a server and a first terminal according to
whether the first terminal supports AI upscaling, according to
embodiments of the disclosure;
FIG. 17 is a diagram for describing a method of performing
streaming between a server and a first terminal according to
whether the first terminal supports AI upscaling, according to
embodiments of the disclosure;
FIG. 18 is a diagram for describing a method, performed by the
server, of streaming image data according to a capability of the
terminal, according to embodiments of the disclosure;
FIG. 19 is a diagram for describing a method, performed by the
server, of streaming image data according to a state of a network
and a capability of the terminal, according to embodiments of the
disclosure;
FIG. 20 is a diagram for describing a method, performed by the
terminal, of streaming image data corresponding to a state of a
network, based on additional information and a capability,
according to embodiments of the disclosure;
FIG. 21 is a diagram for describing additional information provided
for streaming, according to embodiments of the disclosure;
FIG. 22 is a diagram for describing detail configuration of
additional information, according to embodiments of the
disclosure;
FIG. 23 is a diagram for describing detail configuration of
additional information, according to embodiments of the
disclosure;
FIG. 24 is a diagram for describing detail configuration of
additional information, according to embodiments of the
disclosure;
FIG. 25 is a diagram for describing AI data and image data that are
streamed from a server to a terminal, according to embodiments of
the disclosure;
FIG. 26 is a diagram for describing a streaming system, according
to embodiments of the disclosure;
FIG. 27 is a block diagram illustrating a configuration of a
server, according to embodiments of the disclosure; and
FIG. 28 is a block diagram illustrating a configuration of a
terminal, according to embodiments of the disclosure.
DETAILED DESCRIPTION
As the disclosure allows for various changes and numerous examples,
particular embodiments will be illustrated in the drawings and
described in detail in the written description. However, this is
not intended to limit the disclosure to particular modes of
practice, and it will be understood that all changes, equivalents,
and substitutes that do not depart from the spirit and technical
scope of the disclosure are encompassed in the disclosure.
In the description of embodiments, detailed explanations of related
art are omitted when it is deemed that they may unnecessarily
obscure the essence of the disclosure. Also, numbers (for example,
a first, a second, and the like) used in the description of the
specification are identifier codes for distinguishing one element
from another.
Also, in the present specification, it will be understood that when
elements are "connected" or "coupled" to each other, the elements
may be directly connected or coupled to each other, but may
alternatively be connected or coupled to each other with an
intervening element therebetween, unless specified otherwise.
In the present specification, regarding an element represented as a
"unit" or a "module", two or more elements may be combined into one
element or one element may be divided into two or more elements
according to subdivided functions. In addition, each element
described hereinafter may additionally perform some or all of
functions performed by another element, in addition to main
functions of itself, and some of the main functions of each element
may be performed entirely by another component.
Also, in the present specification, an `image` or a `picture` may
denote a still image, a moving image including a plurality of
consecutive still images (or frames), or a video.
Also, in the present specification, a deep neural network (DNN) is
a representative example of an artificial neural network model
simulating brain nerves, and is not limited to an artificial neural
network model using an algorithm.
Also, in the present specification, a `parameter` is a value used
in an operation process of each layer forming a neural network, and
for example, may include a weight used when an input value is
applied to a certain operation expression. Here, the parameter may
be expressed in a matrix form. The parameter is a value set as a
result of training, and may be updated through separate training
data.
Also, in the present specification, a `first DNN` indicates a DNN
used for artificial intelligence (AI) down-scaling an image, and a
`second DNN` indicates a DNN used for AI up-scaling an image.
Also, in the present specification, `DNN setting information`
includes information related to an element constituting a DNN. `DNN
setting information` includes the parameter described above as
information related to the element constituting the DNN. The first
DNN or the second DNN may be set by using the DNN setting
information.
Also, in the present specification, an `original image` denotes an
image to be an object of AI encoding, and a `first image` denotes
an image obtained as a result of performing AI down-scaling on the
original image during an AI encoding process. Also, a `second
image` denotes an image obtained via first decoding during an AI
decoding process, and a `third image` denotes an image obtained by
AI up-scaling the second image during the AI decoding process.
Also, in the present specification, `AI down-scale` denotes a
process of decreasing resolution of an image based on AI, and
`first encoding` denotes an encoding process according to an image
compression method based on frequency transformation. Also, `first
decoding` denotes a decoding process according to an image
reconstruction method based on frequency transformation, and `AI
up-scale` denotes a process of increasing resolution of an image
based on AI.
Expressions such as "at least one of," when preceding a list of
elements, modify the entire list of elements and do not modify the
individual elements of the list. For example, the expression "at
least one of a, b or c" indicates only a, only b, only c, both a
and b, both a and c, both b and c, all of a, b, and c, or
variations thereof.
FIG. 1 is a diagram for describing an AI encoding process and an AI
decoding process, according to embodiments.
As described above, when resolution of an image remarkably
increases, the throughput of information for encoding and decoding
the image is increased, and accordingly, a method for improving
efficiency of encoding and decoding of an image is required.
As shown in FIG. 1, according to embodiments of the disclosure, a
first image 115 is obtained by performing AI down-scaling 110 on an
original image 105 having high resolution. Then, first encoding 120
and first decoding 130 are performed on the first image 115 having
relatively low resolution, and thus a bitrate may be largely
reduced compared to when the first encoding 120 and the first
decoding 130 are performed on the original image 105.
In FIG. 1, the first image 115 is obtained by performing the AI
down-scaling 110 on the original image 105 and the first encoding
120 is performed on the first image 115 during the AI encoding
process, according to embodiments. During the AI decoding process,
AI encoding data including AI data and image data, which are
obtained as a result of AI encoding is received, a second image 135
is obtained via the first decoding 130, and a third image 145 is
obtained by performing AI up-scaling 140 on the second image
135.
Referring to the AI encoding process in detail, when the original
image 105 is received, the AI down-scaling 110 is performed on the
original image 105 to obtain the first image 115 of certain
resolution or certain quality. Here, the AI down-scaling 110 is
performed based on AI, and AI for the AI down-scaling 110 is
trained jointly with AI for the AI up-scaling 140 of the second
image 135. This is because the AI down-scaling 110 and the AI
up-scaling 120 have two competing objectives of scaling-down and
scaling-up an image, and therefore, when the AI for the AI
down-scaling 110 and the AI for the AI up-scaling 140 are
separately trained, a difference between the original image 105
which is an object of AI encoding and the third image 145
reconstructed through AI decoding is increased.
In embodiments of the disclosure, the AI data may be used to
maintain such a joint relationship during the AI encoding process
and the AI decoding process. Accordingly, the AI data obtained
through the AI encoding process may include information indicating
an up-scaling target, and during the AI decoding process, the AI
up-scaling 140 is performed on the second image 135 according to
the up-scaling target verified based on the AI data.
The AI for the AI down-scaling 110 and the AI for the AI up-scaling
140 may be embodied as a DNN. As will be described later with
reference to FIG. 9, because a first DNN and a second DNN are
jointly trained by sharing loss information under a target, an AI
encoding apparatus may provide target information used during joint
training of the first DNN and the second DNN to an AI decoding
apparatus, and the AI decoding apparatus may perform the AI
up-scaling 140 on the second image 135 to target resolution based
on the provided target information.
Regarding the first encoding 120 and the first decoding 130 of FIG.
1, information amount of the first image 115 obtained by performing
AI down-scaling 110 on the original image 105 may be reduced
through the first encoding 120. The first encoding 120 may include
a process of generating prediction data by predicting the first
image 115, a process of generating residual data corresponding to a
difference between the first image 115 and the prediction data, a
process of transforming the residual data of a spatial domain
component to a frequency domain component, a process of quantizing
the residual data transformed to the frequency domain component,
and a process of entropy-encoding the quantized residual data. Such
first encoding 120 may be performed via one of image compression
methods using frequency transformation, such as MPEG-2, H.264
Advanced Video Coding (AVC), MPEG-4, High Efficiency Video Coding
(HEVC), VC-1, VP8, VP9, and AOMedia Video 1 (AV1).
The second image 135 corresponding to the first image 115 may be
reconstructed by performing the first decoding 130 on the image
data. The first decoding 130 may include a process of generating
the quantized residual data by entropy-decoding the image data, a
process of inverse-quantizing the quantized residual data, a
process of transforming the residual data of the frequency domain
component to the spatial domain component, a process of generating
the prediction data, and a process of reconstructing the second
image 135 by using the prediction data and the residual data. Such
first decoding 130 may be performed via an image reconstruction
method corresponding to one of image compression methods using
frequency transformation, such as MPEG-2, H.264 AVC, MPEG-4, HEVC,
VC-1, VP8, VP9, and AV1, which is used in the first encoding
120.
The AI encoding data obtained through the AI encoding process may
include the image data obtained as a result of performing the first
encoding 120 on the first image 115, and the AI data related to the
AI down-scaling 110 of the original image 105. The image data may
be used during the first decoding 130 and the AI data may be used
during the AI up-scaling 140.
The image data may be transmitted in a form of a bitstream. The
image data may include data obtained based on pixel values in the
first image 115, for example, residual data that is a difference
between the first image 115 and prediction data of the first image
115. Also, the image data includes information used during the
first encoding 120 performed on the first image 115. For example,
the image data may include prediction mode information, motion
information, and information related to quantization parameter used
during the first encoding 120. The image data may be generated
according to a rule, for example, according to a syntax, of an
image compression method used during the first encoding 120, among
MPEG-2, H.264 AVC, MPEG-4, HEVC, VC-1, VP8, VP9, and AV1.
The AI data is used in the AI up-scaling 140 based on the second
DNN. As described above, because the first DNN and the second DNN
are jointly trained, the AI data includes information enabling the
AI up-scaling 140 to be performed accurately on the second image
135 through the second DNN. During the AI decoding process, the AI
up-scaling 140 may be performed on the second image 135 to have
targeted resolution and/or quality, based on the AI data.
The AI data may be transmitted together with the image data in a
form of a bitstream. Alternatively, according to embodiments, the
AI data may be transmitted separately from the image data, in a
form of a frame or a packet. The AI data and the image data
obtained as a result of the AI encoding may be transmitted through
the same network or through different networks.
FIG. 2 is a block diagram of a configuration of an AI decoding
apparatus 100 according to embodiments.
Referring to FIG. 2, the AI decoding apparatus 200 according to
embodiments may include a receiver 210 and an AI decoder 230. The
receiver 210 may include a communication interface 212, a parser
214, and an output interface 216. The AI decoder 230 may include a
first decoder 232 and an AI up-scaler 234.
The receiver 210 receives and parses AI encoding data obtained as a
result of AI encoding, and distinguishably outputs image data and
AI data to the AI decoder 230.
The communication interface 212 receives the AI encoding data
obtained as the result of AI encoding through a network. The AI
encoding data obtained as the result of performing AI encoding
includes the image data and the AI data. The image data and the AI
data may be received through a same type of network or different
types of networks.
The parser 214 receives the AI encoding data received through the
communication interface 212 and parses the AI encoding data to
distinguish the image data and the AI data. For example, the parser
214 may distinguish the image data and the AI data by reading a
header of data obtained from the communication interface 212.
According to embodiments, the parser 214 distinguishably transmits
the image data and the AI data to the output interface 216 via the
header of the data received through the communication interface
212, and the output interface 216 transmits the distinguished image
data and AI data respectively to the first decoder 232 and the AI
up-scaler 234. At this time, it may be verified that the image data
included in the AI encoding data is image data generated via a
codec (for example, MPEG-2, H.264 AVC, MPEG-4, HEVC, VC-1, VP8,
VP9, or AV1). In this case, corresponding information may be
transmitted to the first decoder 232 through the output interface
216 such that the image data is processed via the verified
codec.
According to embodiments, the AI encoding data parsed by the parser
214 may be obtained from a data storage medium including a magnetic
medium such as a hard disk, a floppy disk, or a magnetic tape, an
optical recording medium such as CD-ROM or DVD, or a
magneto-optical medium such as a floptical disk.
The first decoder 232 reconstructs the second image 135
corresponding to the first image 115, based on the image data. The
second image 135 obtained by the first decoder 232 is provided to
the AI up-scaler 234. According to embodiments, first decoding
related information, such as prediction mode information, motion
information, quantization parameter information, or the like
included in the image data may be further provided to the AI
up-scaler 234.
Upon receiving the AI data, the AI up-scaler 234 performs AI
up-scaling on the second image 135, based on the AI data. According
to embodiments, the AI up-scaling may be performed by further using
the first decoding related information, such as the prediction mode
information, the quantization parameter information, or the like
included in the image data.
The receiver 210 and the AI decoder 230 according to embodiments
are described as individual devices, but may be implemented through
one processor. In this case, the receiver 210 and the AI decoder
230 may be implemented through an dedicated processor or through a
combination of software and general-purpose processor such as
application processor (AP), central processing unit (CPU) or
graphic processing unit (GPU). The dedicated processor may be
implemented by including a memory for implementing embodiments of
the disclosure or by including a memory processor for using an
external memory.
Also, the receiver 210 and the AI decoder 230 may be configured by
a plurality of processors. In this case, the receiver 210 and the
AI decoder 230 may be implemented through a combination of
dedicated processors or through a combination of software and
general-purpose processors such as AP, CPU or GPU. Similarly, the
AI up-scaler 234 and the first decoder 232 may be implemented by
different processors.
The AI data provided to the AI up-scaler 234 includes information
enabling the second image 135 to be processed via AI up-scaling.
Here, an up-scaling target should correspond to down-scaling of a
first DNN. Accordingly, the AI data includes information for
verifying a down-scaling target of the first DNN.
Examples of the information included in the AI data include
difference information between resolution of the original image 105
and resolution of the first image 115, and information related to
the first image 115.
The difference information may be expressed as information about a
resolution conversion degree of the first image 115 compared to the
original image 105 (for example, resolution conversion rate
information). Also, because the resolution of the first image 115
is verified through the resolution of the reconstructed second
image 135 and the resolution conversion degree is verified
accordingly, the difference information may be expressed only as
resolution information of the original image 105. Here, the
resolution information may be expressed as vertical/horizontal
sizes or as a ratio (16:9, 4:3, or the like) and a size of one
axis. Also, when there is pre-set resolution information, the
resolution information may be expressed in a form of an index or
flag.
The information related to the first image 115 may include
information about either one or both of a bitrate of the image data
obtained as the result of performing first encoding on the first
image 115, and a codec type used during the first encoding of the
first image 115.
The AI up-scaler 234 may determine the up-scaling target of the
second image 135, based on either one or both of the difference
information, and the information related to the first image 115,
which are included in the AI data. The up-scaling target may
indicate, for example, to what degree resolution is to be up-scaled
for the second image 135. When the up-scaling target is determined,
the AI up-scaler 234 performs AI up-scaling on the second image 135
through a second DNN to obtain the third image 145 corresponding to
the up-scaling target.
Before describing a method, performed by the AI up-scaler 234, of
performing AI up-scaling on the second image 135 according to the
up-scaling target, an AI up-scaling process through the second DNN
will be described with reference to FIGS. 3 and 4.
FIG. 3 is a diagram showing a second DNN 300 for performing AI
up-scaling on the second image 135, and FIG. 4 is a diagram for
describing a convolution operation in a first convolution layer 310
of FIG. 3.
As shown in FIG. 3, the second image 135 is input to the first
convolution layer 310. 3.times.3.times.4 indicated in the first
convolution layer 310 shown in FIG. 3 indicates that a convolution
process is performed on one input image by using four filter
kernels having a size of 3.times.3. Four feature maps are generated
by the four filter kernels as a result of the convolution process.
Each feature map indicates inherent characteristics of the second
image 135. For example, each feature map may represent a vertical
direction characteristic, a horizontal direction characteristic, or
an edge characteristic, etc. of the second image 135.
A convolution operation in the first convolution layer 310 will be
described in detail with reference to FIG. 4.
One feature map 450 may be generated through multiplication and
addition between parameters of a filter kernel 430 having a size of
3.times.3 used in the first convolution layer 310 and corresponding
pixel values in the second image 135. Four filter kernels are used
in the first convolution layer 310, and four feature maps may be
generated through the convolution operation using the four filter
kernels.
I1 through I49 indicated in the second image 135 in FIG. 4 indicate
pixels in the second image 135, and F1 through F9 indicated in the
filter kernel 430 indicate parameters of the filter kernel 430.
Also, M1 through M9 indicated in the feature map 450 indicate
samples of the feature map 450.
In FIG. 4, the second image 135 includes 49 pixels, but the number
of pixels is only an example and when the second image 135 has a
resolution of 4 K, the second image 135 may include, for example,
3840.times.2160 pixels.
During a convolution operation process, pixel values of I1, I2, I3,
I8, I9, I10, I15, I16, and I17 of the second image 135 and F1
through F9 of the filter kernels 430 are respectively multiplied,
and a value of combination (for example, addition) of result values
of the multiplication may be assigned as a value of M1 of the
feature map 450. When a stride of the convolution operation is 2,
pixel values of I3, I4, I5, I10, I11, I12, I17, I18, and I19 of the
second image 135 and F1 through F9 of the filter kernels 430 are
respectively multiplied, and the value of the combination of the
result values of the multiplication may be assigned as a value of
M2 of the feature map 450.
While the filter kernel 430 moves along the stride to the last
pixel of the second image 135, the convolution operation is
performed between the pixel values in the second image 135 and the
parameters of the filter kernel 430, and thus the feature map 450
having a certain size may be generated.
According to the present disclosure, values of parameters of a
second DNN, for example, values of parameters of a filter kernel
used in convolution layers of the second DNN (for example, F1
through F9 of the filter kernel 430), may be optimized through
joint training of a first DNN and the second DNN. As described
above, the AI up-scaler 234 may determine an up-scaling target
corresponding to a down-scaling target of the first DNN based on AI
data, and determine parameters corresponding to the determined
up-scaling target as the parameters of the filter kernel used in
the convolution layers of the second DNN.
Convolution layers included in the first DNN and the second DNN may
perform processes according to the convolution operation process
described with reference to FIG. 4, but the convolution operation
process described with reference to FIG. 4 is only an example and
is not limited thereto.
Referring back to FIG. 3, the feature maps output from the first
convolution layer 310 may be input to a first activation layer
320.
The first activation layer 320 may assign a non-linear feature to
each feature map. The first activation layer 320 may include a
sigmoid function, a Tan h function, a rectified linear unit (ReLU)
function, or the like, but is not limited thereto.
The first activation layer 320 assigning the non-linear feature
indicates that at least one sample value of the feature map, which
is an output of the first convolution layer 310, is changed. Here,
the change is performed by applying the non-linear feature.
The first activation layer 320 determines whether to transmit
sample values of the feature maps output from the first convolution
layer 310 to a second convolution layer 330. For example, some of
the sample values of the feature maps are activated by the first
activation layer 320 and transmitted to the second convolution
layer 330, and some of the sample values are deactivated by the
first activation layer 320 and not transmitted to the second
convolution layer 330. The intrinsic characteristics of the second
image 135 represented by the feature maps are emphasized by the
first activation layer 320.
Feature maps 325 output from the first activation layer 320 are
input to the second convolution layer 330. One of the feature maps
325 shown in FIG. 3 is a result of processing the feature map 450
described with reference to FIG. 4 in the first activation layer
320.
3.times.3.times.4 indicated in the second convolution layer 330
indicates that a convolution process is performed on the feature
maps 325 by using four filter kernels having a size of 3.times.3.
An output of the second convolution layer 330 is input to a second
activation layer 340. The second activation layer 340 may assign a
non-linear feature to input data.
Feature maps 345 output from the second activation layer 340 are
input to a third convolution layer 350. 3.times.3.times.1 indicated
in the third convolution layer 350 shown in FIG. 3 indicates that a
convolution process is performed to generate one output image by
using one filter kernel having a size of 3.times.3. The third
convolution layer 350 is a layer for outputting a final image and
generates one output by using one filter kernel. According to
embodiments of the disclosure, the third convolution layer 350 may
output the third image 145 as a result of a convolution
operation.
There may be a plurality of pieces of DNN setting information
indicating the numbers of filter kernels of the first, second, and
third convolution layers 310, 330, and 350 of the second DNN 300, a
parameter of filter kernels of the first, second, and third
convolution layers 310, 330, and 350 of the second DNN 300, and the
like, as will be described later, and the plurality of pieces of
DNN setting information may be connected to a plurality of pieces
of DNN setting information of a first DNN. The connection between
the plurality of pieces of DNN setting information of the second
DNN and the plurality of pieces of DNN setting information of the
first DNN may be realized via joint training of the first DNN and
the second DNN.
In FIG. 3, the second DNN 300 includes three convolution layers
(the first, second, and third convolution layers 310, 330, and 350)
and two activation layers (the first and second activation layers
320 and 340), but this is only an example, and the numbers of
convolution layers and activation layers may vary according to
embodiments. Also, according to embodiments, the second DNN 300 may
be implemented as a recurrent neural network (RNN). In this case, a
convolutional neural network (CNN) structure of the second DNN 300
according to embodiments of the disclosure is changed to an RNN
structure.
According to embodiments, the AI up-scaler 234 may include at least
one arithmetic logic unit (ALU) for the convolution operation and
the operation of the activation layer described above. The ALU may
be implemented as a processor. For the convolution operation, the
ALU may include a multiplier that performs multiplication between
sample values of the second image 135 or the feature map output
from previous layer and sample values of the filter kernel, and an
adder that adds result values of the multiplication. Also, for the
operation of the activation layer, the ALU may include a multiplier
that multiplies an input sample value by a weight used in a
pre-determined sigmoid function, a Tan h function, or an ReLU
function, and a comparator that compares a multiplication result
and a certain value to determine whether to transmit the input
sample value to a next layer.
Hereinafter, a method, performed by the AI up-scaler 234, of
performing the AI up-scaling on the second image 135 according to
the up-scaling target will be described.
According to embodiments, the AI up-scaler 234 may store a
plurality of pieces of DNN setting information settable in a second
DNN.
Here, the DNN setting information may include information about any
one or any combination of the number of convolution layers included
in the second DNN, the number of filter kernels for each
convolution layer, and a parameter of each filter kernel. The
plurality of pieces of DNN setting information may respectively
correspond to various up-scaling targets, and the second DNN may
operate based on DNN setting information corresponding to an
up-scaling target. The second DNN may have different structures
based on the DNN setting information. For example, the second DNN
may include three convolution layers based on any piece of DNN
setting information, and may include four convolution layers based
on another piece of DNN setting information.
According to embodiments, the DNN setting information may only
include a parameter of a filter kernel used in the second DNN. In
this case, the structure of the second DNN does not change, but
only the parameter of the internal filter kernel may change based
on the DNN setting information.
The AI up-scaler 234 may obtain the DNN setting information for
performing AI up-scaling on the second image 135, among the
plurality of pieces of DNN setting information. Each of the
plurality of pieces of DNN setting information used at this time is
information for obtaining the third image 145 of pre-determined
resolution and/or pre-determined quality, and is trained jointly
with a first DNN.
For example, one piece of DNN setting information among the
plurality of pieces of DNN setting information may include
information for obtaining the third image 145 of resolution twice
higher than resolution of the second image 135, for example, the
third image 145 of 4 K (4096.times.2160) twice higher than 2 K
(2048.times.1080) of the second image 135, and another piece of DNN
setting information may include information for obtaining the third
image 145 of resolution four times higher than the resolution of
the second image 135, for example, the third image 145 of 8 K
(8192.times.4320) four times higher than 2 K (2048.times.1080) of
the second image 135.
Each of the plurality of pieces of DNN setting information is
obtained jointly with DNN setting information of the first DNN of
an AI encoding apparatus 600 of FIG. 6, and the AI up-scaler 234
obtains one piece of DNN setting information among the plurality of
pieces of DNN setting information according to an enlargement ratio
corresponding to a reduction ratio of the DNN setting information
of the first DNN. In this regard, the AI up-scaler 234 may verify
information of the first DNN. In order for the AI up-scaler 234 to
verify the information of the first DNN, the AI decoding apparatus
200 according to embodiments receives AI data including the
information of the first DNN from the AI encoding apparatus
600.
In other words, the AI up-scaler 234 may verify information
targeted by DNN setting information of the first DNN used to obtain
the first image 115 and obtain the DNN setting information of the
second DNN trained jointly with the DNN setting information of the
first DNN, by using information received from the AI encoding
apparatus 600.
When DNN setting information for performing the AI up-scaling on
the second image 135 is obtained from among the plurality of pieces
of DNN setting information, input data may be processed based on
the second DNN operating according to the obtained DNN setting
information.
For example, when any one piece of DNN setting information is
obtained, the number of filter kernels included in each of the
first, second, and third convolution layers 310, 330, and 350 of
the second DNN 300 of FIG. 3, and the parameters of the filter
kernels are set to values included in the obtained DNN setting
information.
Parameters of a filter kernel of 3.times.3 used in any one
convolution layer of the second DNN of FIG. 4 are set to {1, 1, 1,
1, 1, 1, 1, 1, 1}, and when DNN setting information is changed
afterwards, the parameters are replaced by {2, 2, 2, 2, 2, 2, 2, 2,
2} that are parameters included in the changed DNN setting
information.
The AI up-scaler 234 may obtain the DNN setting information for AI
up-scaling from among the plurality of pieces of DNN setting
information, based on information included in the AI data, and the
AI data used to obtain the DNN setting information will now be
described.
According to embodiments, the AI up-scaler 234 may obtain the DNN
setting information for AI up-scaling from among the plurality of
pieces of DNN setting information, based on difference information
included in the AI data. For example, when it is verified that the
resolution (for example, 4 K (4096.times.2160)) of the original
image 105 is twice higher than the resolution (for example, 2 K
(2048.times.1080)) of the first image 115, based on the difference
information, the AI up-scaler 234 may obtain the DNN setting
information for increasing the resolution of the second image 135
two times.
According to embodiments, the AI up-scaler 234 may obtain the DNN
setting information for AI up-scaling the second image 135 from
among the plurality of pieces of DNN setting information, based on
information related to the first image 115 included in the AI data.
The AI up-scaler 234 may pre-determine a mapping relationship
between image-related information and DNN setting information, and
obtain the DNN setting information mapped to the information
related to the first image 115.
FIG. 5 is a table showing a mapping relationship between several
pieces of image-related information and several pieces of DNN
setting information.
Through embodiments according to FIG. 5, it will be determined that
AI encoding and AI decoding processes according to embodiments of
the disclosure do not only consider a change of resolution. As
shown in FIG. 5, DNN setting information may be selected
considering resolution, such as standard definition (SD), high
definition (HD), or full HD, a bitrate, such as 10 Mbps, 15 Mbps,
or 20 Mbps, and codec information, such as AV1, H.264, or HEVC,
individually or collectively. For such consideration of the
resolution, the bitrate and the codec information, training in
consideration of each element may be jointly performed with
encoding and decoding processes during an AI training process (see
FIG. 9).
Accordingly, when a plurality of pieces of DNN setting information
are provided based on image-related information including a codec
type, resolution of an image, and the like, as shown in FIG. 5
according to training, the DNN setting information for AI
up-scaling the second image 135 may be obtained based on the
information related to the first image 115 received during the AI
decoding process.
In other words, the AI up-scaler 234 is capable of using DNN
setting information according to image-related information by
matching the image-related information at the left of a table of
FIG. 5 and the DNN setting information at the right of the
table.
As shown in FIG. 5, when it is verified, from the information
related to the first image 115, that the resolution of the first
image 115 is SD, a bitrate of image data obtained as a result of
performing first encoding on the first image 115 is 10 Mbps, and
the first encoding is performed on the first image 115 via AV1
codec, the AI up-scaler 234 may use A DNN setting information among
the plurality of pieces of DNN setting information.
Also, when it is verified, from the information related to the
first image 115, that the resolution of the first image 115 is HD,
the bitrate of the image data obtained as the result of performing
the first encoding is 15 Mbps, and the first encoding is performed
via H.264 codec, the AI up-scaler 234 may use B DNN setting
information among the plurality of pieces of DNN setting
information.
Also, when it is verified, from the information related to the
first image 115, that the resolution of the first image 115 is full
HD, the bitrate of the image data obtained as the result of
performing the first encoding is 20 Mbps, and the first encoding is
performed via HEVC codec, the AI up-scaler 234 may use C DNN
setting information among the plurality of pieces of DNN setting
information, and when it is verified that the resolution of the
first image 115 is full HD, the bitrate of the image data obtained
as the result of performing the first encoding is 15 Mbps, and the
first encoding is performed via HEVC codec, the AI up-scaler 234
may use D DNN setting information among the plurality of pieces of
DNN setting information. One of the C DNN setting information and
the D DNN setting information is selected based on whether the
bitrate of the image data obtained as the result of performing the
first encoding on the first image 115 is 20 Mbps or 15 Mbps. The
different bitrates of the image data, obtained when the first
encoding is performed on the first image 115 of the same resolution
via the same codec, indicates different qualities of reconstructed
images. Accordingly, a first DNN and a second DNN may be jointly
trained based on an image quality, and accordingly, the AI
up-scaler 234 may obtain DNN setting information according to a
bitrate of image data indicating the quality of the second image
135.
According to embodiments, the AI up-scaler 234 may obtain the DNN
setting information for performing AI up-scaling on the second
image 135 from among the plurality of pieces of DNN setting
information considering both information (prediction mode
information, motion information, quantization parameter
information, and the like) provided from the first decoder 232 and
the information related to the first image 115 included in the AI
data. For example, the AI up-scaler 234 may receive quantization
parameter information used during a first encoding process of the
first image 115 from the first decoder 232, verify a bitrate of
image data obtained as an encoding result of the first image 115
from AI data, and obtain DNN setting information corresponding to
the quantization parameter information and the bitrate. Even when
the bitrates are the same, the quality of reconstructed images may
vary according to the complexity of an image. A bitrate is a value
representing the entire first image 115 on which first encoding is
performed, and the quality of each frame may vary even within the
first image 115. Accordingly, DNN setting information more suitable
for the second image 135 may be obtained when prediction mode
information, motion information, and/or a quantization parameter
obtainable for each frame from the first decoder 232 are/is
considered together, compared to when only the AI data is used.
Also, according to embodiments, the AI data may include an
identifier of mutually agreed DNN setting information. An
identifier of DNN setting information is information for
distinguishing a pair of pieces of DNN setting information jointly
trained between the first DNN and the second DNN, such that AI
up-scaling is performed on the second image 135 to the up-scaling
target corresponding to the down-scaling target of the first DNN.
The AI up-scaler 234 may perform AI up-scaling on the second image
135 by using the DNN setting information corresponding to the
identifier of the DNN setting information, after obtaining the
identifier of the DNN setting information included in the AI data.
For example, identifiers indicating each of the plurality of DNN
setting information settable in the first DNN and identifiers
indicating each of the plurality of DNN setting information
settable in the second DNN may be previously designated. In this
case, the same identifier may be designated for a pair of DNN
setting information settable in each of the first DNN and the
second DNN. The AI data may include an identifier of DNN setting
information set in the first DNN for AI down-scaling of the
original image 105. The AI up-scaler 234 that receives the AI data
may perform AI up-scaling on the second image 135 by using the DNN
setting information indicated by the identifier included in the AI
data among the plurality of DNN setting information.
Also, according to embodiments, the AI data may include the DNN
setting information. The AI up-scaler 234 may perform AI up-scaling
on the second image 135 by using the DNN setting information after
obtaining the DNN setting information included in the AI data.
According to embodiments, when pieces of information (for example,
the number of convolution layers, the number of filter kernels for
each convolution layer, a parameter of each filter kernel, and the
like) constituting the DNN setting information are stored in a form
of a lookup table, the AI up-scaler 234 may obtain the DNN setting
information by combining some values selected from values in the
lookup table, based on information included in the AI data, and
perform AI up-scaling on the second image 135 by using the obtained
DNN setting information.
According to embodiments, when a structure of DNN corresponding to
the up-scaling target is determined, the AI up-scaler 234 may
obtain the DNN setting information, for example, parameters of a
filter kernel, corresponding to the determined structure of
DNN.
The AI up-scaler 234 obtains the DNN setting information of the
second DNN through the AI data including information related to the
first DNN, and performs AI up-scaling on the second image 135
through the second DNN set based on the obtained DNN setting
information, and in this case, memory usage and throughput may be
reduced compared to when features of the second image 135 are
directly analyzed for up-scaling.
According to embodiments, when the second image 135 includes a
plurality of frames, the AI up-scaler 234 may independently obtain
DNN setting information for a certain number of frames, or may
obtain common DNN setting information for entire frames.
FIG. 6 is a diagram showing the second image 135 including a
plurality of frames.
As shown in FIG. 6, the second image 135 may include frames t0
through tn.
According to embodiments, the AI up-scaler 234 may obtain DNN
setting information of a second DNN through AI data, and perform AI
up-scaling on the frames t0 through tn based on the obtained DNN
setting information. In other words, the frames t0 through tn may
be processed via AI up-scaling based on common DNN setting
information.
According to embodiments, the AI up-scaler 234 may perform AI
up-scaling on some of the frames t0 through tn, for example, the
frames t0 through ta, by using `A` DNN setting information obtained
from AI data, and perform AI up-scaling on the frames ta+1 through
tb by using `B` DNN setting information obtained from the AI data.
Also, the AI up-scaler 234 may perform AI up-scaling on the frames
tb+1 through tn by using `C` DNN setting information obtained from
the AI data. In other words, the AI up-scaler 234 may independently
obtain DNN setting information for each group including a number of
frames among the plurality of frames, and perform AI up-scaling on
frames included in each group by using the independently obtained
DNN setting information.
According to embodiments, the AI up-scaler 234 may independently
obtain DNN setting information for each frame forming the second
image 135. In other words, when the second image 135 includes three
frames, the AI up-scaler 234 may perform AI up-scaling on a first
frame by using DNN setting information obtained in relation to the
first frame, perform AI up-scaling on a second frame by using DNN
setting information obtained in relation to the second frame, and
perform AI up-scaling on a third frame by using DNN setting
information obtained in relation to the third frame. DNN setting
information may be independently obtained for each frame included
in the second image 135, according to a method of obtaining DNN
setting information based on information (prediction mode
information, motion information, quantization parameter
information, or the like) provided from the first decoder 232 and
information related to the first image 115 included in the AI data
described above. This is because the mode information, the
quantization parameter information, or the like may be determined
independently for each frame included in the second image 135.
According to embodiments, the AI data may include information about
to which frame DNN setting information obtained based on the AI
data is valid. For example, when the AI data includes information
indicating that DNN setting information is valid up to the frame
ta, the AI up-scaler 234 performs AI up-scaling on the frames t0
through ta by using DNN setting information obtained based on the
AI data. Also, when another piece of AI data includes information
indicating that DNN setting information is valid up to the frame
tn, the AI up-scaler 234 performs AI up-scaling on the frames ta+1
through tn by using DNN setting information obtained based on the
other piece of AI data.
Hereinafter, the AI encoding apparatus 600 for performing AI
encoding on the original image 105 will be described with reference
to FIG. 7.
FIG. 7 is a block diagram of a configuration of the AI encoding
apparatus 600 according to embodiments.
Referring to FIG. 7, the AI encoding apparatus 600 may include an
AI encoder 610 and a transmitter 630. The AI encoder 610 may
include an AI down-scaler 612 and a first encoder 614. The
transmitter 630 may include a data processor 632 and a
communication interface 634.
In FIG. 7, the AI encoder 610 and the transmitter 630 are
illustrated as separate devices, but the AI encoder 610 and the
transmitter 630 may be implemented through one processor. In this
case, the AI encoder 610 and the transmitter 630 may be implemented
through an dedicated processor or through a combination of software
and general-purpose processor such as AP, CPU or graphics
processing unit GPU. The dedicated processor may be implemented by
including a memory for implementing embodiments of the disclosure
or by including a memory processor for using an external
memory.
Also, the AI encoder 610 and the transmitter 630 may be configured
by a plurality of processors. In this case, the AI encoder 610 and
the transmitter 630 may be implemented through a combination of
dedicated processors or through a combination of software and a
plurality of general-purpose processors such as AP, CPU or GPU. The
AI down-scaler 612 and the first encoder 614 may be implemented
through different processors.
The AI encoder 610 performs AI down-scaling on the original image
105 and first encoding on the first image 115, and transmits AI
data and image data to the transmitter 630. The transmitter 630
transmits the AI data and the image data to the AI decoding
apparatus 200.
The image data includes data obtained as a result of performing the
first encoding on the first image 115. The image data may include
data obtained based on pixel values in the first image 115, for
example, residual data that is a difference between the first image
115 and prediction data of the first image 115. Also, the image
data includes information used during a first encoding process of
the first image 115. For example, the image data may include
prediction mode information, motion information, quantization
parameter information used to perform the first encoding on the
first image 115, and the like.
The AI data includes information enabling AI up-scaling to be
performed on the second image 135 to an up-scaling target
corresponding to a down-scaling target of a first DNN. According to
embodiments, the AI data may include difference information between
the original image 105 and the first image 115. Also, the AI data
may include information related to the first image 115. The
information related to the first image 115 may include information
about any one or any combination of resolution of the first image
115, a bitrate of the image data obtained as the result of
performing the first encoding on the first image 115, and a codec
type used during the first encoding of the first image 115.
According to embodiments, the AI data may include an identifier of
mutually agreed DNN setting information such that the AI up-scaling
is performed on the second image 135 to the up-scaling target
corresponding to the down-scaling target of the first DNN.
Also, according to embodiments, the AI data may include DNN setting
information settable in a second DNN.
The AI down-scaler 612 may obtain the first image 115 obtained by
performing the AI down-scaling on the original image 105 through
the first DNN. The AI down-scaler 612 may determine the
down-scaling target of the original image 105, based on a
pre-determined standard.
In order to obtain the first image 115 matching the down-scaling
target, the AI down-scaler 612 may store a plurality of pieces of
DNN setting information settable in the first DNN. The AI
down-scaler 612 obtains DNN setting information corresponding to
the down-scaling target from among the plurality of pieces of DNN
setting information, and performs the AI down-scaling on the
original image 105 through the first DNN set in the obtained DNN
setting information.
Each of the plurality of pieces of DNN setting information may be
trained to obtain the first image 115 of pre-determined resolution
and/or pre-determined quality. For example, any one piece of DNN
setting information among the plurality of pieces of DNN setting
information may include information for obtaining the first image
115 of resolution half resolution of the original image 105, for
example, the first image 115 of 2 K (2048.times.1080) half 4 K
(4096.times.2160) of the original image 105, and another piece of
DNN setting information may include information for obtaining the
first image 115 of resolution quarter resolution of the original
image 105, for example, the first image 115 of 2 K
(2048.times.1080) quarter 8 K (8192.times.4320) of the original
image 105.
According to embodiments, when pieces of information (for example,
the number of convolution layers, the number of filter kernels for
each convolution layer, a parameter of each filter kernel, and the
like) constituting the DNN setting information are stored in a form
of a lookup table, the AI down-scaler 612 may obtain the DNN
setting information by combining some values selected from values
in the lookup table, based on the down-scaling target, and perform
AI down-scaling on the original image 105 by using the obtained DNN
setting information.
According to embodiments, the AI down-scaler 612 may determine a
structure of DNN corresponding to the down-scaling target, and
obtain DNN setting information corresponding to the determined
structure of DNN, for example, obtain parameters of a filter
kernel.
The plurality of pieces of DNN setting information for performing
the AI down-scaling on the original image 105 may have an optimized
value as the first DNN and the second DNN are jointly trained.
Here, each piece of DNN setting information includes any one or any
combination of the number of convolution layers included in the
first DNN, the number of filter kernels for each convolution layer,
and a parameter of each filter kernel.
The AI down-scaler 612 may set the first DNN with the DNN setting
information obtained for performing the AI down-scaling on the
original image 105 to obtain the first image 115 of a certain
resolution and/or a certain quality through the first DNN. When the
DNN setting information for performing the AI down-scaling on the
original image 105 is obtained from the plurality of pieces of DNN
setting information, each layer in the first DNN may process input
data based on information included in the DNN setting
information.
Hereinafter, a method, performed by the AI down-scaler 612, of
determining the down-scaling target will be described. The
down-scaling target may indicate, for example, by how much is
resolution decreased from the original image 105 to obtain the
first image 115.
According to embodiments, the AI down-scaler 612 may determine the
down-scaling target based on any one or any combination of a
compression ratio (for example, a resolution difference between the
original image 105 and the first image 115, target bitrate, or the
like), compression quality (for example, type of bitrate),
compression history information, and a type of the original image
105.
For example, the AI down-scaler 612 may determine the down-scaling
target based on the compression ratio, the compression quality, or
the like, which is pre-set or input from a user.
As another example, the AI down-scaler 612 may determine the
down-scaling target by using the compression history information
stored in the AI encoding apparatus 600. For example, according to
the compression history information usable by the AI encoding
apparatus 600, encoding quality, a compression ratio, or the like
preferred by the user may be determined, and the down-scaling
target may be determined according to the encoding quality
determined based on the compression history information. For
example, the resolution, quality, or the like of the first image
115 may be determined according to the encoding quality that has
been used most often according to the compression history
information.
As another example, the AI down-scaler 612 may determine the
down-scaling target based on the encoding quality that has been
used more frequently than a certain threshold value (for example,
average quality of the encoding quality that has been used more
frequently than the threshold value), according to the compression
history information.
As another example, the AI down-scaler 612 may determine the
down-scaling target, based on the resolution, type (for example, a
file format), or the like of the original image 105.
According to embodiments, when the original image 105 includes a
plurality of frames, the AI down-scaler 612 may independently
determine down-scaling target for a certain number of frames, or
may determine down-scaling target for entire frames.
According to embodiments, the AI down-scaler 612 may divide the
frames included in the original image 105 into a certain number of
groups, and independently determine the down-scaling target for
each group. The same or different down-scaling targets may be
determined for each group. The number of frames included in the
groups may be the same or different according to the each
group.
According to embodiments, the AI down-scaler 612 may independently
determine a down-scaling target for each frame included in the
original image 105. The same or different down-scaling targets may
be determined for each frame.
Hereinafter, an example of a structure of a first DNN 700 on which
AI down-scaling is based will be described.
FIG. 8 is a diagram showing the first DNN 700 for performing AI
down-scaling on the original image 105.
As shown in FIG. 8, the original image 105 is input to a first
convolution layer 710. The first convolution layer 710 performs a
convolution process on the original image 105 by using 32 filter
kernels having a size of 5.times.5. 32 feature maps generated as a
result of the convolution process are input to a first activation
layer 720. The first activation layer 720 may assign a non-linear
feature to the 32 feature maps.
The first activation layer 720 determines whether to transmit
sample values of the feature maps output from the first convolution
layer 710 to a second convolution layer 730. For example, some of
the sample values of the feature maps are activated by the first
activation layer 720 and transmitted to the second convolution
layer 730, and some of the sample values are deactivated by the
first activation layer 720 and not transmitted to the second
convolution layer 730. Information represented by the feature maps
output from the first convolution layer 710 is emphasized by the
first activation layer 720.
An output 725 of the first activation layer 720 is input to a
second convolution layer 730. The second convolution layer 730
performs a convolution process on input data by using 32 filter
kernels having a size of 5.times.5. 32 feature maps output as a
result of the convolution process are input to a second activation
layer 740, and the second activation layer 740 may assign a
non-linear feature to the 32 feature maps.
An output 745 of the second activation layer 740 is input to a
third convolution layer 750. The third convolution layer 750
performs a convolution process on input data by using one filter
kernel having a size of 5.times.5. As a result of the convolution
process, one image may be output from the third convolution layer
750. The third convolution layer 750 generates one output by using
the one filter kernel as a layer for outputting a final image.
According to embodiments of the disclosure, the third convolution
layer 750 may output the first image 115 as a result of a
convolution operation.
There may be a plurality of pieces of DNN setting information
indicating the numbers of filter kernels of the first, second, and
third convolution layers 710, 730, and 750 of the first DNN 700, a
parameter of each filter kernel of the first, second, and third
convolution layers 710, 730, and 750 of the first DNN 700, and the
like, and the plurality of pieces of DNN setting information may be
connected to a plurality of pieces of DNN setting information of a
second DNN. The connection between the plurality of pieces of DNN
setting information of the first DNN and the plurality of pieces of
DNN setting information of the second DNN may be realized via joint
training of the first DNN and the second DNN.
In FIG. 8, the first DNN 700 includes three convolution layers (the
first, second, and third convolution layers 710, 730, and 750) and
two activation layers (the first and second activation layers 720
and 740), but this is only an example, and the numbers of
convolution layers and activation layers may vary according to
embodiments. Also, according to embodiments, the first DNN 700 may
be implemented as an RNN. In this case, a CNN structure of the
first DNN 700 according to embodiments of the disclosure is changed
to an RNN structure.
According to embodiments, the AI down-scaler 612 may include at
least one ALU for the convolution operation and the operation of
the activation layer described above. The ALU may be implemented as
a processor. For the convolution operation, the ALU may include a
multiplier that performs multiplication between sample values of
the original image 105 or the feature map output from previous
layer and sample values of the filter kernel, and an adder that
adds result values of the multiplication. Also, for the operation
of the activation layer, the ALU may include a multiplier that
multiplies an input sample value by a weight used in a
pre-determined sigmoid function, a Tan h function, or an ReLU
function, and a comparator that compares a multiplication result
and a certain value to determine whether to transmit the input
sample value to a next layer.
Referring back to FIG. 7, upon receiving the first image 115 from
the AI down-scaler 612, the first encoder 614 may reduce an
information amount of the first image 115 by performing first
encoding on the first image 115. The image data corresponding to
the first image 115 may be obtained as a result of performing the
first encoding by the first encoder 614.
The data processor 632 processes either one or both of the AI data
or the image data to be transmitted in a certain form. For example,
when the AI data and the image data are to be transmitted in a form
of a bitstream, the data processor 632 may process the AI data to
be expressed in a form of a bitstream, and transmit the image data
and the AI data in a form of one bitstream through the
communication interface 634. As another example, the data processor
632 may process the AI data to be expressed in a form of bitstream,
and transmit each of a bitstream corresponding to the AI data and a
bitstream corresponding to the image data through the communication
interface 634. As another example, the data processor 632 may
process the AI data to be expressed in a form of a frame or packet,
and transmit the image data in a form of a bitstream and the AI
data in a form of a frame or packet through the communication
interface 634.
The communication interface 634 transmits AI encoding data obtained
as a result of performing AI encoding, through a network. The AI
encoding data obtained as the result of performing AI encoding
includes the image data and the AI data. The image data and the AI
data may be transmitted through a same type of network or different
types of networks.
According to embodiments, the AI encoding data obtained as a result
of processes of the data processor 632 may be stored in a data
storage medium including a magnetic medium such as a hard disk, a
floppy disk, or a magnetic tape, an optical recording medium such
as CD-ROM or DVD, or a magneto-optical medium such as a floptical
disk.
Hereinafter, a method of jointly training the first DNN 700 and the
second DNN 300 will be described with reference to FIG. 9.
FIG. 9 is a diagram for describing a method of training the first
DNN 700 and the second DNN 300.
In embodiments, the original image 105 on which AI encoding is
performed is reconstructed to the third image 145 via an AI
decoding process, and to maintain similarity between the original
image 105 and the third image 145 obtained as a result of AI
decoding, connectivity is established between the AI encoding
process and the AI decoding process. In other words, information
lost in the AI encoding process is reconstructed during the AI
decoding process, and in this regard, the first DNN 700 and the
second DNN 300 are jointly trained.
For accurate AI decoding, ultimately, quality loss information 830
corresponding to a result of comparing a third training image 804
and an original training image 801 shown in FIG. 9 may be reduced.
Accordingly, the quality loss information 830 is used to train both
of the first DNN 700 and the second DNN 300.
First, a training process shown in FIG. 9 will be described.
In FIG. 9, the original training image 801 is an image on which AI
down-scaling is to be performed and a first training image 802 is
an image obtained by performing AI down-scaling on the original
training image 801. Also, the third training image 804 is an image
obtained by performing AI up-scaling on the first training image
802.
The original training image 801 includes a still image or a moving
image including a plurality of frames. According to embodiments,
the original training image 801 may include a luminance image
extracted from the still image or the moving image including the
plurality of frames. Also, according to embodiments, the original
training image 801 may include a patch image extracted from the
still image or the moving image including the plurality of frames.
When the original training image 801 includes the plurality of
frames, the first training image 802, the second training image,
and the third training image 804 also each include a plurality of
frames. When the plurality of frames of the original training image
801 are sequentially input to the first DNN 700, the plurality of
frames of the first training image 802, the second training image
and the third training image 804 may be sequentially obtained
through the first DNN 700 and the second DNN 300.
For joint training of the first DNN 700 and the second DNN 300, the
original training image 801 is input to the first DNN 700. The
original training image 801 input to the first DNN 700 is output as
the first training image 802 via the AI down-scaling, and the first
training image 802 is input to the second DNN 300. The third
training image 804 is output as a result of performing the AI
up-scaling on the first training image 802.
Referring to FIG. 9, the first training image 802 is input to the
second DNN 300, and according to embodiments, a second training
image obtained as first encoding and first decoding are performed
on the first training image 802 may be input to the second DNN 300.
In order to input the second training image to the second DNN 300,
any one codec among MPEG-2, H.264, MPEG-4, HEVC, VC-1, VP8, VP9,
and AV1 may be used. Any one codec among MPEG-2, H.264, MPEG-4,
HEVC, VC-1, VP8, VP9, and AV1 may be used to perform first encoding
on the first training image 802 and first decoding on image data
corresponding to the first training image 802.
Referring to FIG. 9, separate from the first training image 802
being output through the first DNN 700, a reduced training image
803 obtained by performing legacy down-scaling on the original
training image 801 is obtained. Here, the legacy down-scaling may
include any one or any combination of bilinear scaling, bicubic
scaling, lanczos scaling, or stair step scaling.
To prevent a structural feature of the first image 115 from
deviating greatly from a structural feature of the original image
105, the reduced training image 803 is obtained to preserve the
structural feature of the original training image 801.
Before training is performed, the first DNN 700 and the second DNN
300 may be set to pre-determined DNN setting information. When the
training is performed, structural loss information 810, complexity
loss information 820, and the quality loss information 830 may be
determined.
The structural loss information 810 may be determined based on a
result of comparing the reduced training image 803 and the first
training image 802. For example, the structural loss information
810 may correspond to a difference between structural information
of the reduced training image 803 and structural information of the
first training image 802. Structural information may include
various features extractable from an image, such as luminance,
contrast, histogram, or the like of the image. The structural loss
information 810 indicates how much structural information of the
original training image 801 is maintained in the first training
image 802. When the structural loss information 810 is small, the
structural information of the first training image 802 is similar
to the structural information of the original training image
801.
The complexity loss information 820 may be determined based on
spatial complexity of the first training image 802. For example, a
total variance value of the first training image 802 may be used as
the spatial complexity. The complexity loss information 820 is
related to a bitrate of image data obtained by performing first
encoding on the first training image 802. It is defined that the
bitrate of the image data is low when the complexity loss
information 820 is small.
The quality loss information 830 may be determined based on a
result of comparing the original training image 801 and the third
training image 804. The quality loss information 830 may include
any one or any combination of an L1-norm value, an L2-norm value,
an Structural Similarity (SSIM) value, a Peak Signal-To-Noise
Ratio-Human Vision System (PSNR-HVS) value, an Multiscale SSIM
(MS-SSIM) value, a Variance Inflation Factor (VIF) value, and a
Video Multimethod Assessment Fusion (VMAF) value regarding the
difference between the original training image 801 and the third
training image 804. The quality loss information 830 indicates how
similar the third training image 804 is to the original training
image 801. The third training image 804 is more similar to the
original training image 801 when the quality loss information 830
is small.
Referring to FIG. 9, the structural loss information 810, the
complexity loss information 820 and the quality loss information
830 are used to train the first DNN 700, and the quality loss
information 830 is used to train the second DNN 300. In other
words, the quality loss information 830 is used to train both the
first and second DNNs 700 and 300.
The first DNN 700 may update a parameter such that final loss
information determined based on the structural loss information
810, the complexity loss information 820, and the quality loss
information 830 is reduced or minimized. Also, the second DNN 300
may update a parameter such that the quality loss information 830
is reduced or minimized.
The final loss information for training the first DNN 700 and the
second DNN 300 may be determined as Equation 1 below.
LossDS=a.times.Structural loss information+b.times.Complexity loss
information+c.times.Quality loss information LossUS=d.times.Quality
loss information [Equation 1]
In Equation 1, LossDS indicates final loss information to be
reduced or minimized to train the first DNN 700, and LossUS
indicates final loss information to be reduced or minimized to
train the second DNN 300. Also, a, b, c and d may be predetermined
weights.
In other words, the first DNN 700 updates parameters in a direction
LossDS of Equation 1 is reduced, and the second DNN 300 updates
parameters in a direction LossUS is reduced. When the parameters of
the first DNN 700 are updated according to LossDS derived during
the training, the first training image 802 obtained based on the
updated parameters becomes different from a previous first training
image 802 obtained based on not updated parameters, and
accordingly, the third training image 804 also becomes different
from a previous third training image 804. When the third training
image 804 becomes different from the previous third training image
804, the quality loss information 830 is also newly determined, and
the second DNN 300 updates the parameters accordingly. When the
quality loss information 830 is newly determined, LossDS is also
newly determined, and the first DNN 700 updates the parameters
according to newly determined LossDS. In other words, updating of
the parameters of the first DNN 700 leads to updating of the
parameters of the second DNN 300, and updating of the parameters of
the second DNN 300 leads to updating of the parameters of the first
DNN 700. In other words, because the first DNN 700 and the second
DNN 300 are jointly trained by sharing the quality loss information
830, the parameters of the first DNN 700 and the parameters of the
second DNN 300 may be jointly optimized.
Referring to Equation 1, it is verified that LossUS is determined
according to the quality loss information 830, but this is only an
example and LossUS may be determined based on any one or any
combination of the structural loss information 810 and the
complexity loss information 820, and the quality loss information
830.
Hereinabove, it has been described that the AI up-scaler 234 of the
AI decoding apparatus 200 and the AI down-scaler 612 of the AI
encoding apparatus 600 store the plurality of pieces of DNN setting
information, and methods of training each of the plurality of
pieces of DNN setting information stored in the AI up-scaler 234
and the AI down-scaler 612 will now be described.
As described with reference to Equation 1, the first DNN 700
updates the parameters based on the similarity (the structural loss
information 810) between the structural information of the first
training image 802 and the structural information of the original
training image 801, the bitrate (the complexity loss information
820) of the image data obtained as a result of performing first
encoding on the first training image 802, and the difference (the
quality loss information 830) between the third training image 804
and the original training image 801.
The parameters of the first DNN 700 may be updated such that the
first training image 802 having similar structural information as
the original training image 801 is obtained and the image data
having a small bitrate is obtained when first encoding is performed
on the first training image 802, and at the same time, the second
DNN 300 performing AI up-scaling on the first training image 802
obtains the third training image 804 similar to the original
training image 801.
A direction in which the parameters of the first DNN 700 are
optimized may vary by adjusting the weights a, b, and c of Equation
1. For example, when the weight b is determined to be high, the
parameters of the first DNN 700 may be updated by prioritizing a
low bitrate over high quality of the third training image 804.
Also, when the weight c is determined to be high, the parameters of
the first DNN 700 may be updated by prioritizing high quality of
the third training image 804 over a high bitrate or maintaining of
the structural information of the original training image 801.
Also, the direction in which the parameters of the first DNN 700
are optimized may vary according to a type of codec used to perform
first encoding on the first training image 802. This is because the
second training image to be input to the second DNN 300 may vary
according to the type of codec.
In other words, the parameters of the first DNN 700 and the
parameters of the second DNN 300 may be jointly updated based on
the weights a, b, and c, and the type of codec for performing first
encoding on the first training image 802. Accordingly, when the
first DNN 700 and the second DNN 300 are trained after determining
the weights a, b, and c each to a certain value and determining the
type of codec to a certain type, the parameters of the first DNN
700 and the parameters of the second DNN 300 connected and
optimized to each other may be determined.
Also, when the first DNN 700 and the second DNN 300 are trained
after changing the weights a, b, and c, and the type of codec, the
parameters of the first DNN 700 and the parameters of the second
DNN 300 connected and optimized to each other may be determined. In
other words, the plurality of pieces of DNN setting information
jointly trained with each other may be determined in the first DNN
700 and the second DNN 300 when the first DNN 700 and the second
DNN 300 are trained while changing values of the weights a, b, and
c, and the type of codec.
As described above with reference to FIG. 5, the plurality of
pieces of DNN setting information of the first DNN 700 and the
second DNN 300 may be mapped to the information related to the
first image. To set such a mapping relationship, first encoding may
be performed on the first training image 802 output from the first
DNN 700 via a certain codec according to a certain bitrate and the
second training image obtained by performing first decoding on a
bitstream obtained as a result of performing the first encoding may
be input to the second DNN 300. In other words, by training the
first DNN 700 and the second DNN 300 after setting an environment
such that the first encoding is performed on the first training
image 802 of a certain resolution via the certain codec according
to the certain bitrate, a DNN setting information pair mapped to
the resolution of the first training image 802, a type of the codec
used to perform the first encoding on the first training image 802,
and the bitrate of the bitstream obtained as a result of performing
the first encoding on the first training image 802 may be
determined. By variously changing the resolution of the first
training image 802, the type of codec used to perform the first
encoding on the first training image 802 and the bitrate of the
bitstream obtained according to the first encoding of the first
training image 802, the mapping relationships between the plurality
of DNN setting information of the first DNN 700 and the second DNN
300 and the pieces of information related to the first image may be
determined.
FIG. 10 is a diagram for describing training processes of the first
DNN 700 and the second DNN by a training apparatus 1000.
The training of the first DNN 700 and the second DNN 300 described
with reference FIG. 9 may be performed by the training apparatus
1000. The training apparatus 1000 includes the first DNN 700 and
the second DNN 300. The training apparatus 1000 may be, for
example, the AI encoding apparatus 600 or a separate server. The
DNN setting information of the second DNN 300 obtained as the
training result is stored in the AI decoding apparatus 200.
Referring to FIG. 10, the training apparatus 1000 initially sets
the DNN setting information of the first DNN 700 and the second DNN
300, in operations S840 and S845. Accordingly, the first DNN 700
and the second DNN 300 may operate according to pre-determined DNN
setting information. The DNN setting information may include
information about any one or any combination of the number of
convolution layers included in the first DNN 700 and the second DNN
300, the number of filter kernels for each convolution layer, the
size of a filter kernel for each convolution layer, or a parameter
of each filter kernel.
The training apparatus 1000 inputs the original training image 801
into the first DNN 700, in operation S850. The original training
image 801 may include a still image or at least one frame included
in a moving image.
The first DNN 700 processes the original training image 801
according to the initially set DNN setting information and outputs
the first training image 802 obtained by performing AI down-scaling
on the original training image 801, in operation S855. In FIG. 10,
the first training image 802 output from the first DNN 700 is
directly input to the second DNN 300, but the first training image
802 output from the first DNN 700 may be input to the second DNN
300 by the training apparatus 1000. Also, the training apparatus
1000 may perform first encoding and first decoding on the first
training image 802 via a certain codec, and then input the second
training image to the second DNN 300.
The second DNN 300 processes the first training image 802 or the
second training image according to the initially set DNN setting
information and outputs the third training image 804 obtained by
performing AI up-scaling on the first training image 802 or the
second training image, in operation S860.
The training apparatus 1000 calculates the complexity loss
information 820, based on the first training image 802, in
operation S865.
The training apparatus 1000 calculates the structural loss
information 810 by comparing the reduced training image 803 and the
first training image 802, in operation S870.
The training apparatus 1000 calculates the quality loss information
830 by comparing the original training image 801 and the third
training image 804, in operation S875.
The initially set DNN setting information is updated in operation
S880 via a back propagation process based on the final loss
information. The training apparatus 1000 may calculate the final
loss information for training the first DNN 700, based on the
complexity loss information 820, the structural loss information
810, and the quality loss information 830.
The second DNN 300 updates the initially set DNN setting
information in operation S885 via a back propagation process based
on the quality loss information 830 or the final loss information.
The training apparatus 1000 may calculate the final loss
information for training the second DNN 300, based on the quality
loss information 830.
Then, the training apparatus 1000, the first DNN 700, and the
second DNN 300 may repeat operations S850 through S885 until the
final loss information is minimized to update the DNN setting
information. At this time, during each repetition, the first DNN
700 and the second DNN 300 operate according to the DNN setting
information updated in the previous operation.
Table 1 below shows effects when AI encoding and AI decoding are
performed on the original image 105 according to embodiments of the
disclosure and when encoding and decoding are performed on the
original image 105 via HEVC.
TABLE-US-00001 TABLE 1 Information Subjective Image Amount
(Bitrate) Quality Score (Mbps) (VMAF) Frame AI Encoding/ AI
Encoding/ Content Resolution Number HEVC AI Decoding HEVC AI
Decoding Content_01 8K 300 frames 46.3 21.4 94.80 93.54 Content_02
(7680 .times. 4320) 46.3 21.6 98.05 98.98 Content_03 46.3 22.7
96.08 96.00 Content_04 46.1 22.1 86.26 92.00 Content_05 45.4 22.7
93.42 92.98 Content_06 46.3 23.0 95.99 95.61 Average 46.11 22.25
94.10 94.85
As shown in Table 1, despite subjective image quality when AI
encoding and AI decoding are performed on content including 300
frames of 8 K resolution, according to embodiments of the
disclosure, is higher than subjective image quality when encoding
and decoding are performed via HEVC, a bitrate is reduced by at
least 50%.
FIG. 11 is a diagram of a first apparatus 20 for performing AI
down-scaling on the original image 105 and a second apparatus 40
for performing AI up-scaling on the second image 135.
The first apparatus 20 receives the original image 105 and provides
image data 25 and AI data 30 to the second apparatus 40 by using an
AI down-scaler 1124 and a transformation-based encoder 1126.
According to embodiments, the image data 25 corresponds to the
image data of FIG. 1 and the AI data 30 corresponds to the AI data
of FIG. 1. Also, according to embodiments, the transformation-based
encoder 1126 corresponds to the first encoder 614 of FIG. 7 and the
AI down-scaler 1124 corresponds to the AI down-scaler 612 of FIG.
7.
The second apparatus 40 receives the AI data 30 and the image data
25 and obtains the third image 145 by using a transformation-based
decoder 1146 and an AI up-scaler 1144. According to embodiments,
the transformation-based decoder 1146 corresponds to the first
decoder 232 of FIG. 2 and the AI up-scaler 1144 corresponds to the
AI up-scaler 234 of FIG. 2.
According to embodiments, the first apparatus 20 includes a CPU, a
memory, and a computer program including instructions. The computer
program is stored in the memory. According to embodiments, the
first apparatus 20 performs functions to be described with
reference to FIG. 11 according to execution of the computer program
by the CPU. According to embodiments, the functions to be described
with reference to FIG. 11 are performed by a dedicated hardware
chip and/or the CPU.
According to embodiments, the second apparatus 40 includes a CPU, a
memory, and a computer program including instructions. The computer
program is stored in the memory. According to embodiments, the
second apparatus 40 performs functions to be described with
reference to FIG. 11 according to execution of the computer program
by the CPU. According to embodiments, the functions to be described
with reference to FIG. 11 are performed by a dedicated hardware
chip and/or the CPU.
In FIG. 11, a configuration controller 1122 of the first apparatus
20 receives at least one input value 10. According to embodiments,
the at least one input value 10 may include any one or any
combination of a target resolution difference for the AI
down-scaler 1124 and the AI up-scaler 1144, a bitrate of the image
data 25, a bitrate type of the image data 25 (for example, a
variable bitrate type, a constant bitrate type, or an average
bitrate type), and a codec type for the transformation-based
encoder 1126. The at least one input value 10 may include a value
pre-stored in the first apparatus 20 or a value input from a
user.
The configuration controller 1122 controls operations of the AI
down-scaler 1124 and the transformation-based encoder 1126, based
on the received input value 10. According to embodiments, the
configuration controller 1122 obtains DNN setting information for
the AI down-scaler 1124 according to the received input value 10,
and sets the AI down-scaler 1124 with the obtained DNN setting
information. According to embodiments, the configuration controller
1122 may transmit the received input value 10 to the AI down-scaler
1124 and the AI down-scaler 1124 may obtain the DNN setting
information for performing AI down-scaling on the original image
105, based on the received input value 10. According to
embodiments, the configuration controller 1122 may provide, to the
AI down-scaler 1124, additional information, for example, color
format (luminance component, chrominance component, red component,
green component, or blue component) information to which AI
down-scaling is applied and tone mapping information of a high
dynamic range (HDR), together with the input value 10, and the AI
down-scaler 1124 may obtain the DNN setting information considering
the input value 10 and the additional information. According to
embodiments, the configuration controller 1122 transmits at least a
part of the received input value 10 to the transformation-based
encoder 1126 and the transformation-based encoder 1126 performs
first encoding on the first image 115 via a bitrate of a certain
value, a bitrate of a certain type, and a certain codec.
The AI down-scaler 1124 receives the original image 105 and
performs an operation described with reference to any one or any
combination of FIGS. 1, 7, 8, 9, and 10 to obtain the first image
115.
According to embodiments, the AI data 30 is provided to the second
apparatus 40. The AI data 30 may include either one or both of
resolution difference information between the original image 105
and the first image 115, and information related to the first image
115. The resolution difference information may be determined based
on the target resolution difference of the input value 10, and the
information related to the first image 115 may be determined based
on at least one of a target bitrate, the bitrate type, or the codec
type. According to embodiments, the AI data 30 may include
parameters used during the AI up-scaling. The AI data 30 may be
provided from the AI down-scaler 1124 to the second apparatus
40.
The image data 25 is obtained as the original image 105 is
processed by the transformation-based encoder 1126, and is
transmitted to the second apparatus 40. The transformation-based
encoder 1126 may process the first image 115 according to MPEG-2,
H.264 AVC, MPEG-4, HEVC, VC-1, VP8, VP9, or VA1.
A configuration controller 1142 of the second apparatus 40 controls
an operation of the AI up-scaler 1144, based on the AI data 30.
According to embodiments, the configuration controller 1142 obtains
the DNN setting information for the AI up-scaler 1144 according to
the received AI data 30, and sets the AI up-scaler 1144 with the
obtained DNN setting information. According to embodiments, the
configuration controller 1142 may transmit the received AI data 30
to the AI up-scaler 1144 and the AI up-scaler 1144 may obtain the
DNN setting information for performing AI up-scaling on the second
image 135, based on the AI data 30. According to embodiments, the
configuration controller 1142 may provide, to the AI up-scaler
1144, additional information, for example, the color format
(luminance component, chrominance component, red component, green
component, or blue component) information to which AI up-scaling is
applied, and the tone mapping information of HDR, together with the
AI data 30, and the AI up-scaler 1144 may obtain the DNN setting
information considering the AI data 30 and the additional
information.
According to embodiments, the AI up-scaler 1144 may receive the AI
data 30 from the configuration controller 1142, receive at least
one of prediction mode information, motion information, or
quantization parameter information from the transformation-based
decoder 1146, and obtain the DNN setting information based on the
AI data 30 and any one or any combination of the prediction mode
information, the motion information, and the quantization parameter
information.
The transformation-based decoder 1146 may process the image data 25
to reconstruct the second image 135. The transformation-based
decoder 1146 may process the image data 25 according to MPEG-2,
H.264 AVC, MPEG-4, HEVC, VC-1, VP8, VP9, or AV1.
The AI up-scaler 1144 may obtain the third image 145 by performing
AI up-scaling on the second image 135 provided from the
transformation-based decoder 1146, based on the set DNN setting
information.
The AI down-scaler 1124 may include a first DNN and the AI
up-scaler 1144 may include a second DNN, and according to
embodiments, DNN setting information for the first DNN and second
DNN are trained according to the training method described with
reference to FIGS. 9 and 10.
FIG. 12 is a diagram for describing a concept of a streaming system
1200, according to embodiments of the disclosure.
Referring to FIG. 12, the streaming system 1200 may include a
server 1210 and a terminal 1220. However, this is an example, and
elements of the streaming system 1200 are not limited to the server
1210 and the terminal 1220.
The server 1210 may stream image data to the terminal 1220. In the
disclosure, streaming refers to an operation of transmitting and
receiving image data between the server 1210 and the terminal 1220
such that the terminal 1220 may reproduce the image data in real
time. Also, the server 1210 may stream various types of data
including audio data and text data as well as image data to the
terminal 1220, but in the disclosure, a method of streaming image
data according to embodiments of the disclosure will be
described.
For streaming, the server 1210 and the terminal 1220 may be
connected through a network 1230. The server 1210 may stream image
data to the terminal 1220 via the network 1230.
For example, when the terminal 1220 requests the server 1210 for
predefined image data from among a plurality of items of image
data, the predefined image data may be transmitted to the terminal
1220. The plurality of items of image data may be also referred to
as a plurality of versions of an image content (e.g., a movie, a
television content, a video, etc.), or a plurality of different
quality versions of an image content. The predefined image data may
be image data that corresponds to setting by a user. However, the
disclosure is not limited to the example, and thus, in another
example of the disclosure, the predefined image data may be image
data having a quality set as a default when streaming between the
server 1210 and the terminal 1220 starts. While the server 1210 is
streaming the image data, a state of the network 1230 may be
changeable. Information regarding the state of the network 1230 may
be determined according to an amount of traffic in transmission and
reception paths between the server 1210 and the terminal 1220, and
this may be described as a congestion level. However, this is an
example, and the state of the network 1230 is not described only
according to the traffic occurring in the transmission and
reception paths.
To adaptively perform streaming, based on a changeable state of a
network, the server 1210 may adjust either one or both of a bitrate
and a resolution of image data that is to be transmitted from the
server 1210 to the terminal 1220. The server 1210 may store a
plurality of items of image data 10 (e.g., a high-definition (HD)
class IRON MAN movie, a standard-definition (SD) class IRON MAN
movie, a 15-Mbps IRON MAN movie, a 10-Mbps IRON MAN movie, etc.)
for same image content (e.g., an IRON MAN movie), the plurality of
items of image data 10 being obtained from the same image content
by adjusting either one or both of a bitrate and a resolution.
However, this is an example, and parameters for adjusting a quality
of image data may further include a sampling frequency, a frame
rate, a window size (e.g., 1920.times.1080 (1080p) HD,
1280.times.720 (720p) HD, etc.), a video codec (e.g., H.264, H.265,
Advance Video Coding, etc.), a pixel aspect ratio, an audio codec
(e.g., Advance Audio Coding), or the like.
The server 1210 according to embodiments of the disclosure may
store the plurality of items of image data 10 having different
qualities, and the plurality of items of image data 10 may include
either one or both of AI-encoded image data 12 or non-AI-encoded
image data 14. The AI-encoded image data 12 is generated through
the aforementioned AI encoding process, and the AI encoding process
includes a process of performing AI-downscaling on an original
image through the first DNN 700. In this regard, the first DNN 700
is trained jointly with the second DNN 300 of the terminal 1220,
and when the terminal 1220 receives AI-encoded image data, the
terminal 1220 may perform AI-upscaling on the image data through
the second DNN 300. Also, the AI-encoded image data 12 may be
stored together with AI data related to AI-downscaling, and the AI
data may be used in an AI-upscaling process by the terminal
1220.
The server 1210 may provide additional information of the plurality
of items of image data 10 to the terminal 1220 so as to allow the
terminal 1220 to request image data that corresponds to the state
of the network 1230, from among the plurality of items of image
data 10. The additional information may include quality information
and AI scale conversion information about each of the plurality of
items of image data 10.
A quality of each of the plurality of items of image data 10 may be
determined according to a resolution and a bitrate, and the quality
information may include values of a resolution and a bitrate of
each of the plurality of items of image data 10. However, this is
an example, and the quality may be determined according to a
sampling frequency, a frame rate, a window size, a video codec, a
pixel aspect ratio, an audio codec, or the like. The AI scale
conversion information may include information indicating whether
image data is AI-encoded image data, a value of AI scale conversion
level, or the like. In this regard, the AI scale conversion level
is an index indicating a difference between AI-downscaled image
data and original image data, and may be defined with respect to a
resolution, a bitrate, or the like. For example, when resolutions
of 8K, 4K, full HD (FHD), and HD are supported in the streaming
system 1200, a difference between two adjacent resolutions may be
defined as one level interval. In this case, it may be described
that a level difference between 8K and 4K corresponds to one level
interval, and a level difference between 8K and FHD corresponds to
two level intervals. For example, the resolutions of 8K, 4K, full
HD (FHD), and HD may be set to a first resolution level, a second
resolution level, a third resolution level, and a fourth resolution
level, respectively, and a level interval (also referred to as
"resolution level interval") between the resolutions of 8K, 4K,
full HD (FHD), and HD may be determined as a difference between the
resolution levels set to the resolutions of 8K, 4K, full HD (FHD),
and HD. Also, when bitrates of 40 Mbps, 30 Mbps, 20 Mbps, and 10
Mbps are supported in the streaming system 1200, a difference
between two adjacent bitrates may be defined as one level interval.
For example, the bitrates of 40 Mbps, 30 Mbps, 20 Mbps, and 10 Mbps
may be set to a first bitrate level, a second bitrate level, a
third bitrate level, and a fourth bitrate level, respectively, and
a level interval (also referred to as "bitrate level interval")
between the bitrates of 40 Mbps, 30 Mbps, 20 Mbps, and 10 Mbps may
be determined as a difference between the bitrate levels set to the
bitrates of 40 Mbps, 30 Mbps, 20 Mbps, and 10 Mbps. According to
embodiments of the disclosure, a level and a level interval may be
defined based on a combination of the resolution and the bitrate.
That is, a difference between 8K & 40 Mbps and 8K and 30 Mbps
may be defined as one level interval, and a difference between 8K
& 30 Mbps and 4K & 20 Mbps may be defined as two level
intervals. However, this is an example, and the AI scale conversion
level may be determined according to other factors not only the
resolution and the bitrate.
The terminal 1220 according to embodiments of the disclosure may
check resolutions and bitrates of the plurality of items of image
data 10 included in the additional information, and may request the
server 1210 to transmit image data that has been AI encoded at a
particular resolution or bitrate. Also, the terminal 1220 may
determine a resolution and a bitrate of image data to be requested
for the server 1210, according to a state of the network 1230. For
example, while the terminal 1220 receives AI-encoded image data of
FHD and 5 Mbps from the server 1210, when it is confirmed that a
congestion level of a network is improved because a bit error rate
(BER) of the received image data is decreased, the terminal 1220
may request the server 1210 for AI-encoded image data of 4K and 10
Mbps. In embodiments of the disclosure, the terminal 1220 may
request the server 1210 to increase or decrease the resolution
level, the bitrate level, or the level of the combination of the
resolution level and the bitrate level.
However, this is an example, and the terminal 1220 may request
transmission of image data of a particular resolution or a
particular bitrate, and the server 1210 may determine whether to
transmit AI-encoded image data or non-AI-encoded image data.
Embodiments of the disclosure in which the terminal 1220 requests
the server 1210 for image data based on the additional information
will be further described below with reference to FIGS. 18 to
20.
FIG. 13A is a flowchart for describing a method of streaming data,
the method being performed by a server, according to embodiments of
the disclosure.
In operation S1310, the server 1210 may transmit, to a terminal
1220, additional information of a plurality of items of image data
of different qualities.
In response to a request from the terminal 1220, the server 1210
may transmit, to the terminal 1220, the additional information of
the plurality of items of image data. The additional information
may include quality information and AI scale conversion information
about each of the plurality of items of image data. The quality
information may include resolutions and bitrate values of the
plurality of items of image data, respectively, and the AI scale
conversion information may include information indicating whether
image data is AI-encoded image data, a value of an AI scaling
conversion level, or the like. However, this is an example, and a
plurality of pieces of information included in the additional
information will be further described below with reference to FIGS.
21 to 24.
The additional information may be Media Presentation Description
(MPD) according to the Moving Picture Experts Group (MPEG)-Dynamic
Adaptive Streaming over Hyper Text Transfer Protocol (HTTP) (DASH)
standard. However, this is an example, and the additional
information may be provided as a different type of a manifest file
stored in an Extensible Markup Language (XML) format.
In operation S1320, the server 1210 may receive, from the terminal
1220, a request for image data whose quality corresponds to a state
of a network between the terminal and the server 1210, based on the
additional information.
The server 1210 may receive, from the terminal 1220, a request
message requesting image data of a particular quality from among
the plurality of items of image data, and the request message may
include information for specifying the one from among the plurality
of items of image data.
According to embodiments of the disclosure, the request message may
include quality information about the image data requested by the
terminal 1220. For example, the request message may include
information about either one or both of a bitrate or a
resolution.
According to embodiments of the disclosure, the request message may
include quality information and information indicating whether AI
downscaling has been applied. For example, when AI-encoded image
data is requested, the quality information indicates a quality of
the AI-encoded image data. That is, when the AI-encoded image data
is requested, and the quality information thereof indicates FHD and
5 Mbps, the server 1210 may determine that the terminal requests
the AI-encoded image data whose resolution is FHD and bitrate is 5
Mbps, the AI-encoded image data being obtained as a result of
performing AI downscale. According to embodiments of the
disclosure, when non-AI-encoded image data is requested, the
quality information indicates a quality of original image data.
That is, when the non-AI-encoded image data is requested, and the
quality information thereof indicates FHD and 5 Mbps, the server
1210 may determine that the terminal 1220 requests the original
image data whose resolution is FHD and bitrate is 5 Mbps.
The request message according to embodiments of the disclosure may
include capability information of the terminal 1220 and quality
information about image data requested by the terminal 1220. For
example, when the quality information indicates FHD and 20 Mbps,
and the terminal 1220 supports AI upscale, the server 1210 may
determine that the terminal 1220 requests AI-encoded image data
whose resolution is FHD and bitrate is 20 Mbps, the AI-encoded
image data being obtained as a result of performing AI downscaling.
According to embodiments of the disclosure, when the quality
information indicates FHD and 20 Mbps, and the terminal 1220 does
not support AI upscale, the server 1210 may determine that the
terminal 1220 requests original image data whose resolution is FHD
and bitrate is 20 Mbps.
In operation S1330, in response to the request, the server 1210 may
transmit, to the terminal 1220, AI data and the image data that has
been AI encoded through the first DNN trained jointly with the
second DNN of the terminal 1220.
When the terminal 1220 requests the AI-encoded image data, the
server 1210 may transmit, to the terminal 1220, the AI-encoded
image data together with the AI data including information that may
be used to perform AI upscaling on the AI-encoded image data. For
example, the AI data may include information about any one or any
combination of information indicating whether AI downscaling has
been applied, an AI scale conversion level, and DNN configuration
information used in AI upscaling. However, this is an example, and
the AI data may include other information that may be used in
performing AI upscaling.
The server 1210 may transmit the AI-encoded image data in a unit of
a segment to the terminal 1220. The segment may be generated by
partitioning the AI-encoded image data, based on a preset time
unit. However, this is an example, and a transmission unit of the
AI-encoded image data which is transmitted from the server 1210 is
not limited to the unit of the segment.
In operation S1340, when the state of the network between the
terminal 1220 and the server 1210 is changed, the server 1210 may
receive, from the terminal 1220, a request for image data of a
different quality corresponding to the changed state of the
network.
The terminal 1220 may periodically determine a state of the
network. For example, the terminal 1220 may periodically measure a
timestamp at which image data is received, and a BER, and thus may
determine the state of the network. Also, when the state of the
network is changed, the terminal 1220 may change a quality of image
data to be requested for the server 1210. For example, in a case in
which the terminal 1220 requested the server 1210 for image data of
FHD and 5 Mbps at a first time point at which the network is
congested, the terminal 1220 may request the server 1210 for image
data of 4K and 10 Mbps at a second time point after the first time
point, if the congestion is reduced and the condition of the
network is improved at the second time point. The terminal 1220 or
the server 1210 may determine that a network congestion occurs when
any one or any combination of a delay (or a latency), a bit error
rate, a packet loss, and a timeout (e.g., a lost connection) is
observed. For example, when a time it takes for a destination to
receive a packet sent by a sender (i.e., a delay or a latency) is
longer than a threshold delay, the terminal 1220 or the server 1210
may determine that a network congestion occurs. In another example,
when the terminal 1220 experiences buffering longer than a
threshold buffering time while reproducing a video transmitted from
the server 1210, the terminal 1220 or the server 1210 may determine
that a network congestion occurs.
Information included in a request message to be transmitted from
the terminal 1220 to the server 1210 so as to request the image
data of the different quality corresponding to the changed state of
the network may correspond to the descriptions provided with
reference to S1320.
FIG. 13B is a flowchart for describing a method of streaming data,
the method being performed by a terminal 1220, according to
embodiments of the disclosure.
In operation S1315, the terminal 1220 may request a server 1210 for
image data of a quality corresponding to a state of a network, the
image data being from among a plurality of items of image data,
based on additional information. The additional information may
include quality information and AI scale conversion information
about each of the plurality of items of image data stored in the
server 1210. The quality information may include resolutions and
bitrates of the plurality of items of image data, respectively, and
the AI scale conversion information may include information
indicating whether image data is AI-encoded image data, and a value
of a level at which AI downscaling has been performed.
The terminal 1220 may determine the quality of the image data
corresponding to the state of the network of the terminal 1220,
based on the state of the network. The state of the network may be
determined based on a timestamp and a BER of image data received by
the terminal 1220.
The timestamp refers to information indicating an elapse time from
a reference time to a reception time of image data. For example, in
a case in which an average value of the timestamp in a first time
period is 3 ms whereas an average value of the timestamp in a
second time period is 5 ms, the terminal 1220 may determine that
the state of the network in the second time period is congested
compared to the first time period. Also, the BER refers to a ratio
of an error bit number to a total transmission bit number. For
example, when the BER is lower than a preset reference, the
terminal 1220 may determine that the state of the network is not
congested. As another example, in a case in which a value of the
BER is 0.005 in the first time period whereas a value of the BER is
0.01 in the second time period, the terminal 1220 may determine
that the state of the network in the second time period is
congested compared to the first time period. However, this is an
example, and the state of the network may be determined based on
another information.
For example, when the terminal 1220 estimates that the state of the
network between the server 1210 and the terminal 1220 is congested,
the terminal 1220 may select image data of 20 Mbps that is a
relatively low bitrate from among 50 Mbps, 40 Mbps, 30 Mbps, and 20
Mbps that are respective bitrates of the plurality of items of
image data. However, this is an example, and a method of
determining, by the terminal 1220, the quality of the image data
corresponding to the state of the network is not limited to the
example.
The terminal 1220 may transmit, to the server 1210, a request
message requesting the image data of the determined quality.
According to embodiments of the disclosure, the request message may
include quality information about the image data requested by the
terminal 1220. For example, the request message may include
information about at least one of a bitrate or a resolution. The
request message according to embodiments of the disclosure may
include quality information and information indicating whether AI
downscaling has been applied. The request message according to
embodiments of the disclosure may include capability information of
the terminal 1220 and quality information about the image data
requested by the terminal 1220. The capability information may
include information about whether the terminal 1220 supports AI
upscaling.
In operation S1325, when the terminal 1220 receives the image data
corresponding to the request, and AI data, the terminal 1220 may
determine whether to perform AI upscaling on the received image
data, based on the AI data.
According to embodiments of the disclosure, the terminal 1220 may
determine, based on the AI data, whether the received image data
has been AI encoded through the first DNN trained jointly with the
second DNN. The AI data may include information about at least one
of information indicating whether AI downscaling has been applied,
an AI scale conversion level, or DNN configuration information used
in AI upscaling. The DNN configuration information may be provided
as an indicator indicating the number of convolution layers, the
number of filter kernels of each convolution layer, a parameter of
each filter kernel, or the like. However, this is an example, and
the DNN configuration information may be provided as a lookup
table, and as another example, the second DNN may be provided as
the DNN configuration information. However, this is an example, and
the AI data may include other information required for the terminal
1220 to perform AI upscaling.
According to embodiments of the disclosure, when the terminal 1220
included information specifying AI-encoded image data in the
request message for image data in aforementioned operation S1315,
the terminal 1220 may determine that AI downscaling has been
applied to image data that is received in response to the request
message.
In operation S1335, the terminal 1220 may perform AI upscaling on
the image data received through the second DNN trained jointly with
the first DNN, based on a result of determining whether to perform
AI upscaling.
When the terminal 1220 determines that the received image data is
image data that has been AI encoded through the first DNN trained
jointly with the second DNN, the terminal may perform AI upscaling
on the received image data through the second DNN.
According to embodiments of the disclosure, the terminal 1220 may
determine DNN configuration information of the second DNN, based on
at least one of a resolution or a bitrate of the AI-encoded image
data. For example, when the resolution and the bitrate of the
AI-encoded image data are 4K and 10 Mbps, the terminal 1220 may
select DNN configuration information that is optimized to the
resolution and the bitrate and is from among a plurality of pieces
of DNN configuration information. In this regard, the plurality of
pieces of DNN configuration information that are respectively
optimized to the resolutions and the bitrates may be pre-trained in
the terminal 1220, and information thereof may be included in the
AI data as will be described in embodiments below. According to
embodiments of the disclosure, the terminal 1220 may obtain DNN
configuration information that is optimized for performing AI
upscaling on the AI-encoded image data, based on the DNN
configuration information included in the AI data.
The terminal 1220 may perform AI upscaling on the AI-encoded image
data, based on the selected DNN configuration information, through
the second DNN trained jointly with the first DNN.
In operation S1345, when a state of the network is changed, the
terminal 1220 may request the server 1210 for image data of a
different quality corresponding to the changed state of the
network, based on the additional information.
For example, although the terminal 1220 requested AI-encoded image
data of FHD and 5 Mbps in aforementioned operation S1315, when an
interval of timestamps of image data received thereafter becomes
short or a BER is decreased, the terminal 1220 may determine that a
congestion level of the network is alleviated and improved and thus
may request the server 1210 for AI-encoded image data of 4K and a
10-Mbps bitrate.
As another example, although the terminal 1220 requested AI-encoded
image data of 4K and 10 Mbps in aforementioned operation S1315,
when an interval of timestamps of image data received thereafter
becomes long or a BER is increased, the terminal 1220 may determine
that the congestion level of the network deteriorates and thus may
request the server 1210 for AI-encoded image data of FHD and 5
Mbps.
As another example, although the terminal 1220 requested AI-encoded
image data of FHD and 5 Mbps in aforementioned operation S1315,
when an interval of timestamps of image data received thereafter
becomes long or a BER is increased, the terminal 1220 may determine
that the congestion level of the network deteriorates and thus may
request the server 1210 for image data of HD and 1 Mbps. That is,
when a resolution and a bitrate are less than a predetermined
reference, the terminal 1220 may consider a level of image data
that is to be reconstructed by AI upscaling, and thus may request
the server 1210 for image data on which AI downscaling has not been
performed. However, this is an example, and a method, performed by
the terminal 1220, of changing a quality of image data based on a
change in a state of a network is not limited to the aforementioned
example.
Information included in a request message transmitted from the
terminal 1220 to the server 1210 so as to request the image data of
the different quality corresponding to the state of the network may
correspond to the aforementioned descriptions provided with
reference to operation S1315.
FIG. 14A is a flowchart for describing a method of streaming data,
the method being performed by a server, according to embodiments of
the disclosure.
In operation S1410, the server 1210 may receive, from a terminal
1220, a request for one of a plurality of items of image data of
different qualities determined based on additional information of
the plurality of items of image data.
The additional information may be provided to the terminal 1220
from a service server that is independently separate from the
server 1210. However, this is an example, and the additional
information may be provided from the server 1210 to the terminal
1220.
According to embodiments of the disclosure, the server 1210 may
receive, from the terminal 1220, a request for image data whose
quality is set by a user based on the additional information. For
example, when the user of the terminal 1220 selects a quality of
FHD and 5 Mbps, the server 1210 may receive a request for image
data of FHD and 5 Mbps from the terminal 1220. However, this is an
example, and according to embodiments of the disclosure, the server
1210 may receive, from the terminal 1220, a request for image data
of a quality set as a default. For example, in a case in which
streaming of image data starts between the terminal 1220 and the
server 1210, when a state of a network between the terminal 1220
and the server 1210 is not confirmed, the server 1210 may receive,
from the terminal 1220, a request for image data of a lowest
quality from among a plurality of qualities. According to
embodiments of the disclosure, the server 1210 may receive, from
the terminal 1220, a request for image data of a particular quality
(e.g., HD and 4 Mbps) set as a default. According to embodiments of
the disclosure, the server 1210 may receive, from the terminal
1220, a request for image data for which information about whether
AI encoding has been performed is specified, in addition to a
quality.
In operation S1420, in response to the request, the server 1210 may
transmit, to the terminal 1220, AI data and image data that has
been AI encoded through a DNN for downscaling trained jointly with
a DNN for upscaling of the terminal 1220.
When the terminal 1220 requests AI-encoded image data, the server
1210 may transmit, to the terminal 1220, the AI-encoded image data
together with the AI data including information that may be needed
for upscaling the AI-encoded image data. For example, the AI data
may include information about at least one of information
indicating whether AI downscaling has been applied, an AI scale
conversion level, or DNN configuration information used in AI
upscaling. However, this is an example, and the AI data may include
other information that may be needed for the terminal 1220 to
perform AI upscaling.
The server 1210 may transmit the AI-encoded image data in a unit of
a segment to the terminal 1220. The segment may be generated by
partitioning the AI-encoded image data, based on a preset time
unit. However, this is an example, and a transmission unit of the
AI-encoded image data which is transmitted from the server 1210 is
not limited to the unit of the segment.
In operation S1430, according to the state of the network between
the terminal 1220 and the server 1210, the server 1210 may receive,
from the terminal 1220, a request for image data of a different
quality from among the plurality of items of image data, based on
the additional information.
According to embodiments of the disclosure, the request may include
quality information about the image data requested by the terminal
1220. For example, the request may include information about at
least one of a bitrate or a resolution. As another example, the
request may include quality information and information indicating
whether AI downscaling has been applied. As another example, a
request message may include capability information of the terminal
and quality information about the image data requested by the
terminal 1220.
FIG. 14B is a flowchart for describing a method of streaming data,
the method being performed by the terminal 1220, according to
embodiments of the disclosure.
In operation S1405, the terminal 1220 may request particular image
data. The terminal 1220 according to embodiments of the disclosure
may request image data of a particular quality. For example, the
terminal 1220 may request image data of FHD and 5 Mbps. However,
this is an example, and according to embodiments of the disclosure,
the terminal 1220 may request a server 1210 for image data without
specifying a quality thereof.
According to embodiments of the disclosure, the terminal 1220 may
request a server 1210 for image data by specifying information
about whether AI encoding has been performed, in addition to a
quality.
In operation S1415, the terminal 1220 may receive image data
corresponding to the request. The terminal 1220 according to
embodiments of the disclosure may receive additional information
together with the image data corresponding to the request. The
additional information may include quality information, AI scale
conversion information, or the like about a plurality of items of
image data that may be provided from the server 1210 to the
terminal 1220. However, this is an example, and the additional
information may include a plurality of pieces of other information
for identifying the plurality of items of image data,
respectively.
Also, the additional information may include respective uniform
resource locators (URLs) for receiving the plurality of items of
image data that are identifiable based on the quality information,
the AI scale conversion information, or the like.
Additional information that is to be received by the terminal 1220
may be determined, according to capability information of the
terminal 1220 according to embodiments of the disclosure. For
example, when the terminal 1220 is a device that supports AI
decoding, the server 1210 may transmit additional information
including the AI scale conversion information to the terminal 1220,
and when the terminal 1220 is a device that does not support AI
decoding, the server 1210 may transmit additional information not
including the AI scaling conversion information to the terminal
1220. However, this is an example, and in a case in which the
terminal 1220 is a device that does not support AI decoding, even
when the terminal 1220 receives the additional information
including the AI scale conversion information, the terminal 1220
may not interpret but may ignore the additional information. In the
present embodiment of the disclosure, the capability information of
the terminal 1220 may have been previously provided to the server
1210, or may be included in the request for the particular image
data.
The aforementioned embodiment of the disclosure is an example, and
thus the additional information may be provided from the server
1210 to the terminal 1220 after image data is received during a
predetermined period or may be provided from the server 1210 to the
terminal 1220 before the image data is received.
In operation S1425, the terminal 1220 may determine whether a state
of a network is changed.
The state of the network may be determined based on a timestamp and
a BER of the image data received by the terminal 1220. For example,
as a result of determination based on the timestamp, when the
terminal 1220 determines that a time of receiving the image data
from the server 1210 is delayed, the terminal 1220 may determine
that the state of the network is congested. As another example,
when the BER is less than a predetermined reference, the terminal
1220 may determine that the state of the network is not congested.
However, this is an example, and the state of the network may be
determined based on other information.
According to embodiments of the disclosure, when at least one of
the timestamp or the BER of the image data is changed, the terminal
1220 may determine that the state of the network has been changed,
and according to embodiments of the disclosure, when at least one
of the timestamp or the BER exceeds a preset range, or when a
difference differing from a previous measurement value by at least
a predetermined value occurs or the difference is maintained for a
certain time period, the terminal 1220 may determine that the state
of the network has been changed. For example, when a previous BER
is 0.001, and a BER measured thereafter is in a range of between
0.0095 and 0.005, the terminal 1220 determines that the state of
the network is maintained, but when the BER exceeds the
corresponding range, the terminal 1220 may determine that the state
of the network is changed. However, this is an example, and a
reference by which determination with respect to whether the state
of the network is changed is made is not limited to the
aforementioned example.
As a result of the determination, when the terminal 1220 determines
that the state of the network is not changed, the terminal 1220 may
receive image data corresponding to the quality requested in
operation S1405 or corresponding to whether AI downscaling has been
performed.
In operation S1435, the terminal 1220 may change image data to be
requested, based on the additional information. In aforementioned
operation S1425, when the terminal 1220 determines that the state
of the network has been changed, the terminal 1220 may determine
requirable image data (or image data quality settings), based on
the additional information and the capability of the terminal
1220.
To determine the requirable image data (or the image data quality
settings), the terminal 1220 may determine whether the terminal
1220 can support AI decoding. According to embodiments of the
disclosure, the terminal 1220 may determine whether the terminal
1220 can perform AI upscaling. According to embodiments of the
disclosure, the terminal 1220 may determine whether AI-encoded
image data of a corresponding quality can be a type of DNN
configuration information that can be AI upscaled according to a
type of DNN configuration information, based on the quality
corresponding to a changed state of the network, the DNN
configuration information trained in a second DNN of the terminal
1220, jointly with a first DNN of the server 1210. For example,
when the quality corresponding to the changed state of the network
is FHD and 5 Mbps, the terminal 1220 may determine whether the
second DNN has been trained jointly with the first DNN of the
server 1210 so as to AI upscale the AI-encoded image data of FHD
and 5 Mbps.
According to embodiments of the disclosure, in a case of image data
of a same quality, the terminal 1220 may determine whether AI
downscaling has been performed on the image data to be requested,
an AI downscale level, a type of DNN configuration information used
in the AI downscaling, or the like, based on a hardware device
specification of the terminal 1220, a type of codec, or the like.
Information about whether each image data has been AI downscaled,
an AI downscale level of each image data, a type of DNN
configuration information used in the AI downscaling, or the like
may be included in the additional information and provided to the
terminal 1220, and the descriptions therefor will be further
provided below with reference to FIGS. 20 to 23.
FIG. 15A is a flowchart for describing a method of streaming data,
the method being performed by a server, according to embodiments of
the disclosure.
In operation S1510, the server 1210 may transmit, to a terminal
1220, additional information of a plurality of items of image data
of different qualities.
The server 1210 may transmit the additional information of the
plurality of items of image data, in response to a request from the
terminal 1220. However, this is an example, and the server 1210 may
transmit the additional information to the terminal 1220 when a
communication session for streaming image data is established
between the server 1210 and the terminal 1220.
In the present embodiment of the disclosure, the additional
information may correspond to that described with reference to FIG.
13A.
In operation S1520, the server 1210 may receive, from the terminal
1220, a request for image data from among the plurality of items of
image data, based on the additional information.
According to embodiments of the disclosure, the server 1210 may
receive, from the terminal 1220, a request for image data whose
quality is set by a user. For example, when the user of the
terminal 1220 selects a quality of FHD and 5 Mbps, the server 1210
may receive a request for image data of FHD and 5 Mbps from the
terminal 1220. However, this is an example, and according to
embodiments of the disclosure, the server 1210 may receive, from
the terminal 1220, a request for image data of a quality set as a
default. For example, in a case in which streaming of image data
starts between the terminal 1220 and the server 1210, when a state
of a network between the terminal 1220 and the server is not
confirmed, the server 1210 may receive, from the terminal 1220, a
request for image data of a lowest quality from among a plurality
of qualities. According to embodiments of the disclosure, the
server 1210 may receive, from the terminal 1220, a request for
image data of a particular quality (e.g., HD and 4 Mbps) set as a
default. According to embodiments of the disclosure, the server
1210 may receive, from the terminal, a request for image data for
which information about whether AI encoding has been performed is
specified, in addition to a quality.
In operation S1530, in response to the request, the server 1210 may
transmit, to the terminal 1220, AI data and image data that has
been AI encoded through a DNN for downscaling trained jointly with
a DNN for upscaling of the terminal 1220.
When the terminal 1220 requests AI-encoded image data, the server
1210 may transmit, to the terminal 1220, the AI-encoded image data
together with the AI data including information necessary for
upscaling the AI-encoded image data. For example, the AI data may
include information about at least one of information indicating
whether AI downscaling has been applied, an AI scale conversion
level, or DNN configuration information used in AI upscaling.
However, this is an example, and the AI data may include other
information necessary for the terminal 1220 to perform AI
upscaling.
The server 1210 may transmit the AI-encoded image data in a unit of
a segment to the terminal 1220. The segment may be generated by
partitioning the AI-encoded image data, based on a preset time
unit. However, this is merely an example, and a transmission unit
of the AI-encoded image data which is transmitted from the server
1210 is not limited to the unit of the segment.
In operation S1540, according to the state of the network between
the terminal 1220 and the server 1210, the server 1210 may receive,
from the terminal 1220, a request for image data of a different
quality from among the plurality of items of image data, based on
the additional information.
According to embodiments of the disclosure, the request may include
quality information about the image data requested by the terminal
1220. For example, the request may include information about at
least one of a bitrate or a resolution. As another example, the
request may include quality information and information indicating
whether AI downscaling has been applied. As another example, a
request message may include capability information of the terminal
1220 and quality information about the image data requested by the
terminal 1220.
FIG. 15B is a flowchart for describing a method of streaming data,
the method being performed by the terminal 1220, according to
embodiments of the disclosure.
In operation S1515, the terminal 1220 may receive, from a server
1210, additional information of a plurality of items of image data
of different qualities. According to embodiments of the disclosure,
the terminal 1220 may request the server 1210 for the additional
information of the plurality of items of image data. However, this
is an example, and the server 1210 may transmit the additional
information to the terminal 1220 when a communication session for
streaming image data is established between the terminal 1220 and
the server 1210.
In the present embodiment of the disclosure, the additional
information may correspond to that described with reference to FIG.
13B.
In operation S1525, the terminal 1220 may request the server 1210
for image data from among the plurality of items of image data,
based on the additional information. According to embodiments of
the disclosure, the terminal 1220 may request the server 1210 for
image data whose quality is set by a user. For example, when the
user of the terminal 1220 selects a quality of FHD and 5 Mbps, the
terminal 1220 may request the server 1210 for image data of FHD and
5 Mbps from the terminal 1220. However, this is an example, and
according to embodiments of the disclosure, the terminal 1220 may
request the server 1210 for image data of a quality set as a
default. For example, in a case in which streaming of image data
starts between the terminal 1220 and the server 1210, when a state
of a network between the terminal 1220 and the server 1210 is not
confirmed, the terminal 1220 may request the server 1210 for image
data of a lowest quality from among a plurality of qualities.
According to embodiments of the disclosure, the terminal 1220 may
request the server 1210 for image data of a particular quality
(e.g., HD and 4 Mbps) set as a default.
According to embodiments of the disclosure, the terminal 1220 may
request the server 1210 for image data by specifying information
about whether AI encoding has been performed, in addition to a
quality.
In operation S1535, when the terminal 1220 receives image data and
AI data which correspond to the request, the terminal 1220 may
determine whether to perform AI upscaling on the received image
data, based on the AI data.
According to embodiments of the disclosure, the terminal 1220 may
determine, based on the AI data, whether the received image data
has been AI encoded through the first DNN trained jointly with the
second DNN. The AI data may include information about at least one
of information indicating whether AI downscaling has been applied,
an AI scale conversion level, or DNN configuration information used
in AI upscaling.
According to embodiments of the disclosure, when the terminal 1220
includes the information that specifies AI-encoded image data in a
request message for image data in operation S1525, the terminal
1220 may determine that AI downscaling has been applied to image
data received in response to the request message.
In operation S1545, the terminal 1220 may perform, based on a
result of the determination about whether to perform AI upscaling,
AI upscaling on the received image data through the DNN for
upscaling trained jointly with the DNN for downscaling of the
server 1210.
When the terminal 1220 determines that the received image data is
image data that has been AI encoded through the first DNN trained
jointly with the second DNN, the terminal 1220 may perform, through
the second DNN, AI upscaling on the received image data. In the
present embodiment of the disclosure, a method, performed by the
terminal 1220, of performing AI upscaling on received image data
through the second DNN may correspond to operation S1335 described
above with reference to FIG. 13B.
In operation S1555, the terminal 1220 may confirm the state of the
network between the terminal 1220 and the server 1210. The state of
the network may be determined based on a timestamp and a BER of
image data received by the terminal 1220. For example, as a result
of determination based on the timestamp, when the terminal 1220
determines that a time of receiving the image data from the server
1210 is delayed, the terminal 1220 may determine that the state of
the network is congested. As another example, when the BER is less
than a predetermined reference, the terminal 1220 may determine
that the state of the network is not congested. However, this is an
example, and the state of the network may be determined based on
other information.
In operation S1565, according to the state of the network, the
terminal 1220 may request the server 1210 for image data of a
different quality from among the plurality of items of image data,
based on the additional information.
For example, although the terminal 1220 requested AI-encoded image
data of FHD and 5 Mbps in aforementioned operation S1525, when an
interval of timestamps of image data received thereafter becomes
short or a BER is decreased, the terminal 1220 may determine that a
congestion level of the network is improved and thus may request
the server 1210 for AI-encoded image data of 4K and a 10-Mbps
bitrate.
As another example, although the terminal 1220 requested AI-encoded
image data of 4K and 10 Mbps in aforementioned operation S1525,
when an interval of timestamps of image data received thereafter
becomes long or a BER is increased, the terminal 1220 may determine
that the congestion level of the network deteriorates and thus may
request the server 1210 for AI-encoded image data of FHD and 5
Mbps.
As another example, although the terminal 1220 requested AI-encoded
image data of FHD and 5 Mbps in aforementioned operation S1515,
when an interval of timestamps of image data received thereafter
becomes long or a BER is increased, the terminal 1220 may determine
that the congestion level of the network deteriorates and thus may
request the server 1210 for image data of HD and 1 Mbps. That is,
when a resolution and a bitrate are less than a predetermined
reference, the terminal 1220 may consider a level of image data
that is to be reconstructed by AI upscaling, and thus may request
the server 1210 for image data on which AI downscaling has not been
performed. However, this is an example, and a method, performed by
the terminal 1220, of changing a quality of image data based on a
change in a state of a network is not limited to the aforementioned
example.
Information included in a request message transmitted from the
terminal 1220 to the server 1210 so as to request the image data of
the different quality corresponding to the state of the network may
correspond to the aforementioned descriptions provided with
reference to operation S1515.
FIG. 16 is a diagram for describing a method of performing
streaming between a server 1610 and a first terminal 1620 according
to whether the first terminal 1620 supports AI upscaling, according
to embodiments of the disclosure.
In the embodiment of FIG. 16, it is assumed that the first terminal
1620 corresponds to a terminal that can support AI upscaling
through a second DNN trained jointly with a first DNN of the server
1610, and a second terminal 1630 corresponds to a terminal that
does not support AI upscaling.
The server 1610 stores a plurality of items of image data of
different qualities for adaptive streaming, and transmits image
data, in response to a request from a terminal (e.g., the first
terminal 1620). The terminal (e.g., the first terminal 1620) may
obtain additional information of the plurality of items of image
data from the server 1610, and may request the server 1610 for one
of the plurality of items of image data, based on the additional
information. A plurality of pieces of information included in the
additional information will be further described below with
reference to FIGS. 20 to 23.
For example, the server 1610 may obtain image data of 4K and 20
Mbps 1642 that is AI downscaled by performing AI downscaling on
original image data of 8K and 60 Mbps 1640 through a 1a-1 DNN 1612.
The server 1610 may transmit the AI-encoded image data of 4K and 20
Mbps 1642 together with AI data related to AI downscaling to 4K and
20 Mbps, in response to a request from the first terminal 1620. The
first terminal 1620 may perform AI upscaling on the received image
data through a 2a-1 DNN 1622 trained jointly with the 1a-1 DNN
1612, and thus may obtain AI upscaled image data 1652. In this
regard, the first terminal 1620 may perform aforementioned AI
upscaling by obtaining at least one of information indicating
whether AI downscaling has been applied, an AI scale conversion
level, or DNN configuration information used in the AI upscaling,
based on information included in the received AI data.
As another example, the server 1610 may obtain image data of FHD
and 7 Mbps 1644 that is AI downscaled by performing AI downscaling
on the original image data of 8K and 60 Mbps 1640 through a 1a-2
DNN 1614. The server 1610 may transmit the AI-encoded image data of
FHD and 7 Mbps 1644 together with AI data related to AI downscaling
to FHD and 7 Mbps, in response to a request from the first terminal
1620. For example, while the first terminal 1620 requests and
receives the AI-encoded image data of 4K and 20 Mbps 1642 as in the
aforementioned example, when the first terminal 1620 determines
that a congestion level of a network deteriorates, the first
terminal 1620 may request the AI-encoded image data of FHD and 7
Mbps 1644. The first terminal 1620 may perform AI upscaling on the
image data, which is received in response to the request, through a
2a-2 DNN 1624 trained jointly with the 1a-2 DNN 1614, and thus may
obtain AI upscaled image data 1654. As in the aforementioned
example, the first terminal 1620 may use information in AI
upscaling, the information being included in the AI data.
As another example, the server 1610 may obtain image data of HD and
4 Mbps 1646 that is AI downscaled by performing AI downscaling on
the original image data of 8K and 60 Mbps 1640 through a 1a-3 DNN
1616. The server 1610 may transmit the AI-encoded image data of HD
and 4 Mbps 1646 together with AI data related to AI downscaling to
HD and 4 Mbps, in response to a request from the first terminal
1620. The first terminal 1620 may perform AI upscaling on the
received image data through a 2a-3 DNN 1626 trained jointly with
the 1a-3 DNN 1616, and thus may obtain AI upscaled image data 1656.
As in the aforementioned example, the first terminal 1620 may use
information in AI upscaling, the information being included in the
AI data.
Also, the first terminal 1620 may additionally perform legacy
upscaling on AI upscaled image data. For example, due to a state of
a network, the first terminal 1620 may receive the AI-encoded image
data of HD and 4 Mbps 1646 obtained by applying AI downscaling
through the 1a-3 DNN 1616. The first terminal 1620 may obtain the
AI upscaled image data 1656 of FHD and 7 Mbps by performing AI
upscaling through the 2a-3 DNN 1626 trained jointly with the 1 a-3
DNN 1616. The first terminal 1620 may perform legacy upscaling on
the AI upscaled image data 1656 of FHD and 7 Mbps and thus may
obtain image data of 4K and 20 Mbps.
In the aforementioned example, it is described that the first
terminal 1620 receives AI-encoded image data through a first DNN,
but the first terminal 1620 may receive image data that is not AI
downscaled (e.g., image data 1650).
The second terminal 1630 is a terminal that does not support
upscaling through a second DNN trained jointly with the first DNN.
The second terminal 1630 may not determine whether image data 1660,
1662, 1664, or 1666 received from the server 1610 is image data on
which AI downscaling has been performed through the first DNN, and
may process the received image data 1660, 1662, 1664, or 1666. In a
case of image data on which downscaling has been performed through
the first DNN, the image data may have less quality loss compared
to image data on which a general downscale technique has been
performed, and thus, even when the second terminal 1630 does not
support upscaling through the second DNN, the second terminal 1630
may be provided image data of a high quality, compared to image
data based on the related art.
FIG. 17 is a diagram for describing a method of performing
streaming between a server 1710 and a first terminal 1720 according
to whether the first terminal 1720 supports AI upscaling, according
to embodiments of the disclosure.
In the embodiment of FIG. 17, it is assumed that the first terminal
1720 corresponds to a terminal that can support AI upscaling
through a second DNN trained jointly with a first DNN of the server
1710, and a second terminal 1730 corresponds to a terminal that
does not support AI upscaling.
The server 1710 stores a plurality of items of image data of
different qualities for adaptive streaming, and transmits image
data, in response to a request from a terminal (e.g., the first
terminal 1720). The terminal (e.g., the first terminal 1720) may
obtain additional information of the plurality of items of image
data from the server 1710, and may request the server 1710 for one
of the plurality of items of image data, based on the additional
information. A plurality of pieces of information included in the
additional information will be further described below with
reference to FIGS. 21 to 24.
For example, the server 1710 may obtain image data of 4K and 20
Mbps 1742 that is AI downscaled by performing AI downscaling on
original image data of 8K and 60 Mbps 1740 through a 1 b-1 DNN
1712. The server 1710 may transmit the AI-encoded image data of 4K
and 20 Mbps 1742 together with AI data related to AI downscaling to
4K and 20 Mbps, in response to a request from the first terminal
1720. The first terminal 1720 may perform AI upscaling on the
received image data through a 2b-1 DNN 1722 trained jointly with
the 1 b-1 DNN 1712, and thus may obtain AI upscaled image data
1752. In this regard, the first terminal 1720 may perform
aforementioned AI upscaling by obtaining at least one of
information indicating whether AI downscaling has been applied, an
AI scale conversion level, or DNN configuration information used in
the AI upscaling, based on information included in the received AI
data.
As another example, the server 1710 may perform AI downscaling on
the AI downscaled image data of 4K 1742 through a 1b-2 DNN 1714,
and thus may obtain AI downscaled image data of FHD and 7 Mbps
1744. The server 1710 may transmit the AI downscaled image data of
FHD and 7 Mbps 1744 together with AI data related to AI downscaling
to FHD and 7 Mbps, in response to a request from the first terminal
1720. For example, while the first terminal 1720 requests and
receives the AI-encoded image data of 4K and 20 Mbps 1742 as in the
aforementioned example, when the first terminal 1720 determines
that a congestion level of a network deteriorates, the first
terminal 1720 may request the AI-encoded image data of FHD and 7
Mbps 1744. The first terminal 1720 may perform AI upscaling on the
image data, which is received in response to the request, through a
2b-2 DNN 1724 trained jointly with the 1b-2 DNN 1714, and thus may
obtain AI upscaled image data 1754. As in the aforementioned
example, the first terminal 1720 may use information in AI
upscaling, the information being included in the AI data.
As another example, the server 1710 may perform AI downscaling on
the AI-encoded image data of FHD and 7 Mbps 1744 through a 2b-3 DNN
1716, and thus may obtain AI downscaled image data of HD and 4 Mbps
1746. The server 1710 may transmit the AI downscaled image data of
HD and 4 Mbps 1746 together with AI data related to AI downscaling
to HD and 4 Mbps, in response to a request from the first terminal
1720. The first terminal 1720 may perform AI upscaling on the
received image data through a 2b-3 DNN 1726 trained jointly with
the 1b-3 DNN 1716, and thus may obtain AI upscaled image data 1756.
As in the aforementioned example, the first terminal 1720 may use
information in AI upscaling, the information being included in the
AI data.
In FIG. 17, the server 1710 is illustrated as obtaining the AI
downscaled image data of FHD and 7 Mbps 1744 through two separate
downscaling processes using the 1 b-1 DNN 1712 and the 1b-2 DNN
1714, but the embodiments are not limited thereto, and the server
1710 may downscale the original image data of 8K and 60 Mbps 1740
directly to the image data of FHD 1744 using a single DNN, which is
jointly trained with a corresponding DNN of the first terminal
1720. Also, the server 1710 may downscale the original image data
of 8K and 60 Mbps 1740 directly to the image data of HD 1746 using
a single DNN, which is jointly trained with a corresponding DNN of
the first terminal 1720.
In the aforementioned example, it is described that the first
terminal 1720 receives AI-encoded image data through a first DNN,
but the first terminal 1720 may receive image data that is not AI
downscaled (e.g., image data 1750).
A structure of a DNN shown in FIG. 17 is an example, and at least
one of the 1 b-1 DNN 1712, the 1b-2 DNN 1714, or the 1b-3 DNN 1716
may be replaced with a legacy scaler. Also, in association thereto,
at least one of the 2b-1 DNN 1722, the 2b-2 DNN 1724, or the 2b-3
DNN 1726 may be replaced with a legacy scaler.
The second terminal 1730 is a terminal that does not support
upscaling through a second DNN trained jointly with the first DNN.
The second terminal 1730 may not determine whether image data 1760,
1762, 1764, or 1766 received from the server 1710 is image data on
which AI downscaling has been performed through the first DNN, and
may process the received image data 1760, 1762, 1764, or 1766. In a
case of image data on which downscaling has been performed through
the first DNN, the image data may have less quality loss compared
to image data on which a general downscale technique has been
performed, and thus, even when the second terminal 1730 does not
support upscaling through the second DNN, the second terminal 1730
may be provided image data of a high quality, compared to image
data based on the related art.
The image data shown in FIGS. 12 and 17 is an example, and a
plurality of items of image data stored in different qualities are
not limited to AI-encoded 4K image data, AI-encoded FHD image data,
AI-encoded HD image data, or the like. For example, the server 1710
may perform AI downscaling on 8K original image data and thus may
store a plurality of items of AI-encoded image data including 5K
(5120.times.2880) image data, 3K (2560.times.1440) image data, 540p
(960.times.540) image data, 360p (640.times.360) image data, or the
like. Also, the server 1710 may store a plurality of items of image
data of different bitrates that are with respect to image data of a
particular resolution. For example, AI-encoded 4K image data may be
stored as image data of 4K and 20 Mbps, image data of 4K and 15
Mbps, or the like. To this end, the first DNN structure shown in
FIGS. 16 and 17 may be variously configured. That is, a first DNN
structure for converting 8K image data to AI-encoded 3K image data,
a first DNN structure for converting 3K image data to AI-encoded
540p image data, a first DNN structure for converting image data of
8K and 60 Mbps to AI-encoded image data of 4K and 15 Mbps, or the
like may be used.
FIG. 18 is a diagram for describing a method, performed by the
server 1210, of streaming image data according to a capability of
the terminal 1220, according to embodiments of the disclosure.
In operation S1810, the terminal 1220 may transmit information
about the capability to the server 1210. According to embodiments
of the disclosure, the information about the capability may include
at least one of information indicating whether the terminal 1220
can change a quality of image data requested adaptive to a state of
a network, information indicating whether the terminal 1220 can
support AI upscaling through a second DNN, or information about an
AI upscale level supportable by the terminal 1220. However, this is
an example, and the information about the capability may include
information about codec supported by the terminal 1220.
In operation S1820, the terminal 1220 may request the server 1210
for additional information. For adaptive streaming between the
server 1210 and the terminal 1220, it is required to check
respective qualities of a plurality of items of image data
providable from the server 1210 and whether AI encoding has been
performed thereto. Accordingly, the terminal 1220 may request the
server 1210 for the additional information. The additional
information will be further described below with reference to FIGS.
21 to 24. Accordingly, the terminal 1220 may request the server
1210 for additional information of the plurality of items of image
data.
In operation S1830, the server 1210 may transmit the additional
information to the terminal 1220. When the server 1210 receives the
request from the terminal 1220, the server 1210 may determine the
additional information corresponding to the request. For example,
the server 1210 may determine the additional information that is
from among a plurality of pieces of additional information stored
in the server 1210 and is requested by the terminal 1220, based on
an identifier of the additional information included in the request
from the terminal 1220. The additional information may be directly
generated by the server 1210, but according to embodiments of the
disclosure, the additional information may be received from a
different server.
In operation S1840, the terminal 1220 may request the server 1210
for image data of a quality corresponding to a state of a network,
based on the additional information.
In the present embodiment of the disclosure, it is assumed that a
request message transmitted from the terminal 1220 to the server
1210 so as to request image data includes only quality information
corresponding to the state of the network. For example, the
terminal 1220 may transmit, to the server 1210, a request message
including information indicating that a resolution of the image
data corresponding to the state of the network corresponds to FHD,
and a bitrate of the image data corresponds to 5 Mbps.
In operation S1850, the server 1210 may determine image data
corresponding to a request, based on the capability of the terminal
1220.
When image data of a particular quality is requested, the server
1210 may determine whether to transmit image data on which
downscaling has been performed. As a result of determination by the
server 1210 based on the capability of the terminal 1220, when the
terminal 1220 can support AI upscaling through a second DNN trained
jointly with a first DNN, the server 1210 may transmit AI-encoded
image data.
Also, the server 1210 may determine which image data is to be
transmitted, based on an AI upscale level supportable by the
terminal 1220, the image data being AI downscaled to a certain
level. For example, in a case in which the terminal 1220 requested
AI-encoded image data of FHD and 5 Mbps, the server 1210 may
determine whether to transmit image data of FHD and 5 Mbps obtained
by performing AI downscaling on image data of 8K and 30 Mbps
through a 1a DNN, or image data of FHD and 5 Mbps obtained by
performing AI downscaling on image data of 4K and 10 Mbps through a
1 b DNN. The 1 a DNN and the 1 b DNN may have different DNN
configuration information.
Table 1 below includes values of a resolution and a bitrate which
are providable from the server 1210 to the terminal 1220. When
describing a scale conversion level with reference to Table 1, a
difference between 4K & 20 Mbps and 4K and 10 Mbps, and a
difference between 4K & 10 Mbps and FHD and 5 Mbps may each be
defined as one level. However, this is an example, and resolutions
and bitrates that are supported by a streaming system according to
the disclosure are not limited to the values in Table 1.
TABLE-US-00002 TABLE 1 Resolution Bitrate 8K 40 Mbps 30 Mbps 4K 20
Mbps 10 Mbps FHD 5 Mbps HD 1 Mbps
In operation S1860, the server 1210 may transmit AI data and image
data based on the determination to the terminal 1220. The AI data
may include information required for the terminal 1220 to AI
upscale AI-encoded image data, and may correspond to the
descriptions provided with reference to FIG. 5.
In operation S1870, the terminal 1220 may perform AI upscaling on
the received image data through the second DNN trained jointly with
the first DNN.
The terminal 1220 may determine, based on the AI data, whether to
apply AI upscaling to the received image data through the second
DNN trained jointly with the first DNN. When the AI data includes
information indicating that the received image data is AI-encoded
image data, the terminal 1220 may perform AI upscaling through the
second DNN on the received image data. Also, the AI data may
include information about at least one of an AI scale conversion
level or DNN configuration information used in AI upscaling. For
example, the AI data may include information indicating whether the
AI-encoded image data of FHD and 5 Mbps is generated by AI
downscaling the image data of 8K and 30 Mbps through the 1a DNN or
by AI downscaling the image data of 4K and 10 Mbps through the 1 b
DNN.
FIG. 19 is a diagram for describing a method, performed by the
server 1210, of streaming image data according to a state of a
network and a capability of the terminal 1220, according to
embodiments of the disclosure.
In operation S1910, the terminal 1220 may transmit information
about the capability to the server 1210. According to embodiments
of the disclosure, the information about the capability may include
at least one of information indicating whether the terminal 1220
can change a quality of image data requested adaptive to a state of
a network, information indicating whether the terminal 1220 can
support AI upscaling through a second DNN, or information about an
AI upscale level supportable by the terminal 1220. However, this is
an example, and the information about the capability may include
information about codec supported by the terminal 1220.
In operation S1920, the terminal 1220 may request the server 1210
for additional information. The terminal 1220 may request the
server 1210 for additional information of image data. Operation
S1920 may correspond to operation S1820 described above with
reference to FIG. 18.
In operation S1930, the server 1210 may transmit the additional
information to the terminal 1220. The server 1210 may transmit the
additional information, in response to the request from the
terminal 1220. Operation S1930 may correspond to operation S1830
described above with reference to FIG. 18.
In operation S1940, the terminal 1220 may request the server 1210
for image data of a particular quality, based on the additional
information. For example, the terminal 1220 may request the server
1210 for image data of 8K and 30 Mbps, according to selection by a
user. Also, the terminal 1220 may include information about a state
of a network in the request for the image data of a particular
quality, and may transmit the request. For example, the terminal
1220 may include, in a request message, information about a BER, a
timestamp, or the like of previously-received image data, and may
transmit the request message to the server 1210.
In operation S1950, the server 1210 may determine image data
corresponding to the request, based on the state of the network and
the capability of the terminal 1220.
The server 1210 may determine a quality of the image data
corresponding to the state of the network, based on the information
about the state of the network included in the request received
from the terminal 1220. For example, the server 1210 may determine
the quality of the image data to be 4K and 20 Mbps, the quality of
the image data corresponding to the state of the network.
As described above in operation S1940, when it is determined that
the terminal 1220 requested the image data of 8K and 30 Mbps but
the request from the terminal 1220 does not correspond to the state
of the network, the server 1210 may determine whether the terminal
1220 supports AI upscaling through the second DNN trained jointly
with the first DNN, based on the information about the capability
of the terminal 1220. As a result of the determination by the
server 1210, when the terminal 1220 supports AI upscaling, the
server 1210 may determine to transmit, to the terminal 1220, image
data of 4K and 20 Mbps that is generated by performing AI
downscaling on the image data of 8K and 30 Mbps through the first
DNN.
In operation S1960, the server 1210 may transmit AI data and the
image data to the terminal 1220, based on the determination. The AI
data may include information required for the terminal 1220 to AI
upscale AI-encoded image data, and may correspond to the
descriptions provided with reference to FIG. 12.
In operation S1970, the terminal 1220 may perform AI upscaling on
the received image data through the second DNN trained jointly with
the first DNN.
The terminal 1220 may determine, based on the AI data, whether to
apply AI upscaling to the received image data through the second
DNN trained jointly with the first DNN. When the AI data includes
information indicating that the received image data is AI-encoded
image data, the terminal 1220 may perform AI upscaling through the
second DNN on the received image data. Also, the AI data may
include information about at least one of an AI scale conversion
level or DNN configuration information used in AI upscaling. For
example, the AI data may include information indicating that the
AI-encoded image data of 4K and 20 Mbps is generated by AI encoding
the image data of 8K and 30 Mbps. As another example, the AI data
may include an AI scale conversion level, and the terminal 1220 may
determine DNN configuration information, based on the resolution
and the bitrate.
FIG. 20 is a diagram for describing a method, performed by the
terminal 1220, of streaming image data corresponding to a state of
a network, based on additional information and a capability,
according to embodiments of the disclosure.
In operation S2010, the terminal 1220 may transmit information
about the capability to the server 1210. According to embodiments
of the disclosure, the information about the capability may include
at least one of information indicating whether the terminal 1220
can change a quality of image data requested adaptive to the state
of the network, information indicating whether the terminal 1220
can support AI upscaling through a second DNN, or information about
an AI upscale level supportable by the terminal 1220. However, this
is an example, and the information about the capability may include
information about codec supported by the terminal 1220.
In operation S2020, the terminal 1220 may request the server 1210
for the additional information. The terminal 1220 may request the
server 1210 for the additional information of image data. Operation
S2020 may correspond to S1820 described above with reference to
FIG. 18.
In operation S2030, the server 1210 may transmit the additional
information to the terminal 1220. The server 1210 may transmit the
additional information to the terminal 1220, in response to the
request from the terminal 1220. Operation S2030 may correspond to
S1830 described above with reference to FIG. 18.
In operation S2040, the terminal 1220 may identify the image data
that corresponds to the state of the network and is from among a
plurality of items of image data, based on the additional
information and the capability of the terminal 1220.
For example, the terminal 1220 may check, based on the additional
information, respective qualities and whether AI encoding has been
performed about the plurality of items of image data that can be
provided by the server 1210, and various types of DNN configuration
information which can be used in performing AI upscaling on image
data of a particular quality. The additional information will be
described in detail with reference to FIGS. 21 to 24.
The terminal 1220 may determine, based on the capability of the
terminal 1220, the image data that corresponds to the state of the
network and is from among the plurality of items of image data
checked based on the additional information. For example, when a
quality of the image data that corresponds to the state of the
network is FHD and 5 Mbps, the terminal 1220 may determine, based
on the capability of the terminal 1220, one of AI-encoded image
data of FHD and 5 Mbps generated by performing AI downscaling on
image data of 8K and 30 Mbps through the 1a DNN and AI-encoded
image data of FHD and 5 Mbps generated by performing AI downscaling
on image data of 4K and 10 Mbps through the 1 b DNN, wherein the
image data of 8K and 30 Mbps and the image data of 4K and 10 Mbps
are providable from the server 1210. When the terminal 1220
supports AI upscaling through a 2b DNN trained jointly with a 1 b
DNN, the terminal 1220 may determine, from among the plurality of
items of image data, the AI-encoded image data of FHD and 5 Mbps
generated by performing AI downscaling on the image data of 4K and
10 Mbps through the 1b DNN.
In operation S2050, the terminal 1220 may request the server 1210
for the determined image data. A request message transmitted from
the terminal 1220 to the server 1210 so as to request the
determined image data may include an identifier of the determined
image data. For example, the request message may include an
identifier of the AI-encoded image data of FHD and 5 Mbps generated
by performing AI downscaling on the image data of 4K and 10 Mbps
through the 1 b DNN.
The terminal 1220 may request the server 1210 for the determined
image data in a unit of a segment. The segment may be generated by
partitioning the image data, based on a time unit. When the
terminal 1220 requests the determined image data in a unit of a
segment, the request message may include not only information about
a quality of the determined image data, whether AI encoding has
been performed thereon, or the like but may also include an
identifier of the segment. The identifier of the segment may
include a segment number, an offset, or the like, but this is an
example and thus the identifier of the segment is not limited to
the aforementioned examples. The segment number refers to each of
numbers respectively allocated to a plurality of segments included
in the image data. Also, the offset refers to a difference between
a preset reference time and a start time of the segment. Here, the
preset reference time may be a start time of a first segment or an
initialization segment from among the plurality of segments
included in the image data.
In operation S2060, the server 1210 may transmit AI data and the
image data corresponding to a request. The AI data may include
information required for the terminal 1220 to AI upscale the
AI-encoded image data, and may correspond to the descriptions
provided with reference to FIG. 5.
In operation S2070, the terminal 1220 may perform AI upscaling on
the received image data through the second DNN trained jointly with
the first DNN.
The terminal 1220 may perform AI upscaling on the received image
data, based on information about at least one of an AI scale
conversion level or DNN configuration information used in AI
upscaling, which is included in the AI data. A method by which the
terminal 1220 performs AI upscaling on the received image data may
correspond to the descriptions provided with reference to FIG.
2.
FIG. 21 is a diagram for describing additional information provided
for streaming, according to embodiments of the disclosure.
Referring to FIG. 21, the additional information may hierarchically
include an image data set element 2110, an image data element 2120,
and a segment element 2130. Each of the aforementioned elements
2110, 2120, and 2130 may include a plurality of pieces of
information indicating an image data set, image data, and an
attribute of a segment. The image data set may be a group of a
plurality of items of interchangeable image data. For example, the
image data set may be a group of a plurality of items of image data
generated by encoding a first period of content at different
qualities, and may correspond to an adaptation set in the
MPEG-DASH. The segment may be a portion generated by partitioning
the image data, based on a time.
The image data set element 2110 may include information about a
type 2112 and identification (ID) 2114 of each of image data sets.
In this regard, the type 2112 may indicate a type of content
included in the image data set, and examples of the type may
include an image, audio, a text, or the like. The ID 2114 may
include identifiers for identifying the image data sets,
respectively.
The image data element 2120 may include ID 2122, a quality 2124, AI
scale conversion information 2126, or the like of image data. The
ID 2122 may include an identifier for identifying the image data,
and the quality 2124 may include various attributes including a
bitrate, a resolution, or the like. The AI scale conversion
information 2126 may further include information about codec
appropriate for a second DNN used in AI upscaling conversion,
information about AI upscaling levels that are available in
respective conversions of the plurality of items of image data,
information about a parameter set of the second DNN corresponding
to a parameter set used in a first DNN, or the like.
The segment element 2130 may include information about ID 2132, an
offset 2134, or the like of the segment. The ID 2132 may include an
identifier for identifying the segment, and the offset 2134 may
include information about a position of the segment on a timeline.
In a case in which a quality of image data to be received has to be
changed due to a change in a state of a network, the offset 2134
may be used to synchronize image data of a previous quality with
image data of a quality to be newly received. For example, when the
plurality of items of image data consist of segments whose time
offsets are 2 ms, 4 ms, 6 ms, and 8 ms, respectively, a terminal
may receive segments up to 4 ms with respect to image data of 4K
and 10 Mbps, and afterward, when the terminal requests image data
of FHD and 5 Mbps due to a change in the state of the network, the
terminal may process a segment of 6 ms to be reproduced according
to synchronization with the image data of 4K and 10 Mbps.
A structure of the additional information shown in FIG. 21 is an
example, and additional information for adaptive streaming is not
limited thereto. As another example, the AI scale conversion
information 2126 may be included in the segment element 2130. As
another example, the additional information may additionally
include parameter update information by which the terminal can
update the parameter of the second DNN jointly with the parameter
of the first DNN of a server. However, this is an example, and
information for updating the parameter of the second DNN may be
provided to the terminal, separately from the additional
information.
FIG. 22 is a diagram for describing detail configuration of
additional information, according to embodiments of the
disclosure.
Referring to FIG. 22, a "mediadataset" attribute may be defined in
the additional information according to embodiments of the
disclosure. The "mediadataset" attribute is to indicate an
attribute of a media data set consisting of a plurality of items of
media data of different qualities, and may include an "id" element
indicating an identifier of the media data set, a "type" element
indicating a type of content, or the like. The present embodiment
of the disclosure corresponds to a case in which media data is
image data, and the "type" element may be set as a video.
Hereinafter, descriptions will be provided assuming that the media
data is the image data.
The "mediadataset" attribute defines an attribute of each of a
plurality of items of image data of different qualities, and may
include an "id" element indicating an identifier of the media data,
a "resolution" element indicating a resolution, a "bitrate" element
indicating a bitrate, an "AI upscale" element indicating whether it
is required to apply AI upscaling, or the like. A terminal may
check whether image data has been AI downscaled through a first DNN
trained jointly with a second DNN of the terminal, based on the "AI
upscale" element of each image data. The "AI upscale" element may
be included in the aforementioned AI scale conversion
information.
The terminal may check an attribute of each image data included in
additional information, and may request particular image data,
based on the attribute. For example, the terminal may request a
server for AI-encoded image data of 4K and 10 Mbps from among the
plurality of items of image data. In this case, the terminal may
transmit, to the server, a request message including information
about id=2.
When the terminal receives image data corresponding to the request,
the terminal may perform AI upscaling on the received image data
through the second DNN trained jointly with the first DNN of the
server. In this regard, the terminal may obtain DNN configuration
information that is optimized for the second DNN to perform AI
upscaling on the image data, based on information about a
resolution, a bitrate, or the like of the image data. The DNN
configuration information may include information about filter
kernels (e.g., the number of convolution layers, the number of
filter kernels according to each convolution layer, a parameter of
each filter kernel, or the like). For example, the terminal may
include information that has been trained jointly with the first
DNN of the server so as to indicate that upscaling through the
second DNN has to be performed on image data of 4K and 10 Mbps by
using A DNN configuration information.
According to embodiments of the disclosure, the DNN configuration
information that is optimized for the second DNN to perform AI
upscaling may vary according to not only a resolution and a bitrate
of image data but also according to a genre of content consisting
of the plurality of items of image data. For example, the terminal
may include information that has been trained jointly with the
first DNN of the server so as to indicate that, for a sports genre,
upscaling through the second DNN has to be performed on AI-encoded
image data of 4K and 10 Mbps by using DNN configuration information
corresponding to the sports genre and 4K & 10 Mbps, and for a
drama genre, upscaling through the second DNN has to be performed
on AI-encoded image data of 4K and 10 Mbps by using DNN
configuration information corresponding to the drama genre and 4K
& 10 Mbps.
FIG. 23 is a diagram for describing detail configuration of
additional information, according to embodiments of the
disclosure.
Referring to FIG. 23, a "mediadataset" attribute and a "mediadata"
attribute may be defined in the additional information according to
embodiments of the disclosure. In the present embodiment of the
disclosure, descriptions corresponding to embodiments of the
disclosure which is described above with reference to FIG. 22 are
not provided, and an "AI upscalelevel" element different therefrom
will now be described in detail.
The "AI upscalelevel" element included in the "mediadata" attribute
indicates a difference between AI-encoded image data and original
image data.
As described above with reference to FIG. 5, because the AI
encoding process according to embodiments of the disclosure is
performed based on both a resolution and a bitrate, information of
the difference between the AI-encoded image data and the original
image data may be provided. In the embodiment of FIG. 23, image
data of 8K and 40 Mbps corresponds to the original image data. In
the present embodiment of the disclosure, a value of the "AI
upscalelevel" element may be determined based on a difference
between a bitrate and a resolution of the image data of 8K and 40
Mbps (id=n) and a bitrate and a resolution of the AI-encoded image
data.
For example, AI-encoded image data whose id is 1 may have been
generated by performing AI encoding on the original image data,
based on a bitrate of 30 Mbps, and AI-encoded image data whose id
is 2 may have been generated by performing AI encoding on the
original image data, based on a resolution of 4K and a bitrate of
20 Mbps. The value of the "AI upscalelevel" element according to
embodiments of the disclosure is an example of the difference in
resolutions and bitrates of the original image data and the
AI-encoded image data and thus is not limited to the example.
The terminal may select, based on a capability of the terminal,
image data on which the terminal can perform AI upscaling from
among two level AI-encoded image data of 4K and 20 Mbps, and
three-level AI-encoded image data of 4K and 10 Mbps, and may
request the selected image data.
In the embodiment of FIG. 23, image data whose id is n+1 and image
data whose id is n+2 correspond to image data obtained by
performing downscaling using a legacy downscaler.
FIG. 24 is a diagram for describing detail configuration of
additional information, according to embodiments of the
disclosure.
Referring to FIG. 24, a "mediadataset" attribute and a "mediadata"
attribute may be defined in the additional information according to
embodiments of the disclosure. In the present embodiment of the
disclosure, descriptions corresponding to embodiments of the
disclosure which is described above with reference to FIG. 22 are
not provided, and an "AIupscaleparameterset" element different
therefrom will now be described in detail.
The "AIupscaleparameterset" element included in the "mediadata"
attribute may provide information about a plurality of pieces of
various DNN configuration information that are usable in performing
AI upscaling on AI downscaled image data. For example, even for a
plurality of items of AI-encoded image data of a same quality of 4K
and 20 Mbps, the number of DNN convolution layers and a size and
number of filter kernels, which are used in AI upscaling the for
AI-encoded image data, may vary and thus various DNN configuration
information may exist.
When a plurality of items of AI-encoded image data of a same
quality have different DNN configuration information, the terminal
may determine one of the plurality of items of AI-encoded image
data of the same quality, based on the capability of the terminal.
For example, when an AI up-scaler of the terminal includes B
AIupscaleparameterset, the terminal may select, from among the
plurality of items of AI-encoded image data of the same quality,
AI-encoded image data that can be reconstructed in a corresponding
DNN. As another example, when the AI up-scaler of the terminal
includes A=AIupscaleparameterset having a complicated configuration
compared to that of the aforementioned example, the terminal may
reconstruct all of the plurality of items of AI-encoded image data
of the same quality. In this case, the terminal may select, from
among the plurality of items of AI-encoded image data of the same
quality, AI-encoded image data that uses a relatively less network
resource in streaming or that can be reconstructed to a higher
quality by the terminal, according to configuration.
The terminal may obtain at least one piece of DNN configuration
information from among a plurality of pieces of DNN configuration
information, based on a hardware specification of the terminal or
codec. For example, the terminal may obtain DNN configuration
information that corresponds to the terminal and is from among a
plurality of pieces of DNN configuration information that are
applicable to AI upscaling with respect to same AI-encoded image
data of 4K and 20 Mbps. Accordingly, the terminal may request a
server for image data that has been AI encoded based on the DNN
configuration information corresponding to 4K and 20 Mbps.
FIG. 25 is a diagram for describing AI data 2510 and image data
2520 that are streamed from a server to a terminal, according to
embodiments of the disclosure.
Referring to FIG. 25, in response to a request from the terminal,
the server may transmit, to the terminal, the AI data 2510 and the
image data 2520 that correspond to the request.
The AI data 2510 may include information indicating whether AI
downscaling has been performed on the image data 2520. Also, when
the image data 2520 has been AI encoded, the AI data 2510 may
include information about at least one of an AI scale conversion
level or DNN configuration information used in AI upscaling.
According to embodiments of the disclosure, the AI data 2510 may
include the AI scale conversion level, and the DNN configuration
information for AI upscaling may be determined by the terminal
based on a resolution and a bitrate.
The AI data 2510 may correspond to an initialization segment of the
MPEG-DASH, and the terminal may determine whether the image data
2520 has been AI upscaled through a second DNN trained jointly with
a first DNN of the server.
The AI data 2510 may include other information required for the
terminal to decode the image data 2520, in addition to the
aforementioned information. For example, the AI data 2510 may
include information about a type of codec, ID, an offset, or the
like.
The image data 2520 may consist of a plurality of segments 2522 to
2524. The plurality of segments 2522 to 2524 may be generated by
partitioning the image data 2520, based on a time. In response to a
request from the terminal, the server may transmit the image data
2520 in a unit of a segment to the terminal. Accordingly, when a
state of a network between the server and the terminal is changed,
a quality of image data requested for the server by the terminal
may be efficiently changed.
However, configurations of the AI data 2510 and the image data 2520
are an example, and configurations of AI data and image data for
streaming according to embodiments of the disclosure are not
limited thereto.
FIG. 26 is a diagram for describing a streaming system 2600,
according to embodiments of the disclosure.
Referring to FIG. 26, the streaming system 2600 according to
embodiments of the disclosure may include a service server 2610, a
plurality of content servers 2622 and 2624 (also referred to as the
first content server 2622 and the N content server 2624), and a
terminal 2630. However, this is an example, and the streaming
system 2600 may further include additional elements. For example,
the streaming system 2600 may include a service server 2610. The
service server 2610 may be provided in a multiple number. Also, the
present embodiment of the disclosure will now be described with
reference to one terminal 2630, but the service server 2610 and the
plurality of content servers 2622 and 2624 may stream image data to
a plurality of terminals.
The service server 2610 may provide additional information of a
plurality of items of image data to the terminal 2630 so as to
allow the terminal 2630 to request image data that corresponds to a
state of a network and is from among the plurality of items of
image data. The additional information may include respective
qualities of the plurality of items of image data, whether AI
encoding has been performed thereon, various DNN configuration
information that can be used in performing AI upscaling on image
data of a particular quality, or the like. Also, the additional
information may include location information over the network where
the plurality of items of image data are stored. For example, the
additional information may include a uniform resource identifier
(URI) of the first content server 2622 or the N content server
2624. Also, as described above with reference to FIG. 21, when the
plurality of items of image data are partitioned in a unit of a
segment, the additional information may include an URI of each
segment.
The terminal 2630 may request image data that corresponds to the
state of the network and is from among the plurality of items of
image data, based on the additional information. For example,
according to a result of determination based on the state of the
network, when the terminal 2630 determines to request image data of
FHD and 5 Mbps generated by performing AI downscaling, through the
1a DNN, on image data of 8K and 30 Mbps from among the plurality of
items of image data, the terminal 2630 may obtain URI information
of the determined image data, based on the additional information.
In the present embodiment of the disclosure, the image data of FHD
and 5 Mbps generated by performing AI downscaling, through the 1a
DNN, on the image data of 8K and 30 Mbps may be stored in the first
content server 2622. The terminal 2630 may request the first
content server 2622 for the determined image data, based on the URI
information. When a request is received, the first content server
2622 may transmit AI data and image data corresponding to the
request to the terminal 2630. However, this is an example, and the
terminal 2630 may request the first content server 2622 for a
request for the determined image data in a unit of a segment.
The plurality of content servers 2622 and 2624 may provide
information about content to the service server 2610. A content
server (e.g., the content server 2622) may provide information
about content stored in the content server (e.g., the content
server 2622) or information about new added content to the service
server 2610. The service server 2610 may generate or update the
additional information, based on a plurality of pieces of
information about content which are provided from the plurality of
content servers 2622 and 2624.
The streaming system 2600 described above with reference to FIG. 26
is an example, and a system that performs streaming according to
the disclosure is not limited thereto. For example, at least one
cache server to deliver image data and AI data may be further
provided between the terminal 2630 and the plurality of content
servers 2622 and 2624. As another example, the streaming system
2600 may further include a source server for providing content to
each of the plurality of content servers 2622 and 2624.
FIG. 27 is a block diagram illustrating a configuration of a server
2700, according to embodiments of the disclosure.
Referring to FIG. 27, the server 2700 according to embodiments of
the disclosure may include a communication interface 2710, a
processor 2720, and a memory 2730. However, this is an example, and
the server 2700 may additionally further include other elements.
For example, the server 2700 may include a plurality of
processors.
The communication interface 2710 according to embodiments of the
disclosure may provide an interface for communicating with another
device (e.g., a terminal). The communication interface 2710 may
receive a request for additional information or image data from the
terminal. Also, the communication interface 2710 may transmit the
additional information or media to the terminal.
The processor 2720 according to embodiments of the disclosure may
generally control the server 2700 to execute one or more programs
stored in the memory 2730 to perform operations related to an image
encoding apparatus described above with reference to FIGS. 1 to 11,
and operations related to a server described above with reference
to FIGS. 12 to 26.
The memory 2730 according to embodiments of the disclosure may
store various data, programs, or applications for driving and
controlling the server 2700. Each of the one or more programs
stored in the memory 2730 may include one or more instructions.
Each program (one or more instructions) or each application, which
is stored in the memory 2730, may be executed by the processor
2720.
FIG. 28 is a block diagram illustrating a configuration of a
terminal 2800, according to embodiments of the disclosure.
Referring to FIG. 28, the terminal 2800 according to embodiments of
the disclosure may include a communication interface 2810, a
processor 2820, and a memory 2830. However, this is an example, and
the terminal 2800 may additionally further include other elements.
For example, the terminal 2800 may include a plurality of
processors including a central processing unit (CPU), a graphic
processing unit (GPU), a neutral processing unit (NPU), or the
like.
The communication interface 2810 according to embodiments of the
disclosure may provide an interface for communicating with another
device (e.g., a server). The communication interface 2810 may
transmit a request for additional information or image data to a
server. Also, the communication interface 2810 may receive
additional information or media from the server and may output the
received additional information or the media to the processor
2820.
The processor 2820 according to embodiments of the disclosure may
generally control the terminal 2800 to execute one or more programs
stored in the memory 2830 to perform operations related to an image
decoding apparatus described above with reference to FIGS. 1 to 11,
and operations related to a terminal described above with reference
to FIGS. 12 to 26.
For example, the processor 2820 may transmit a request for
additional information of a plurality of items of image data of
different qualities to the server via the communication interface
2810. In response to the request, the processor 2820 may obtain the
additional information from the server via the communication
interface 2810.
The processor 2820 may transmit a request for predefined image data
from among the plurality of items of image data, based on the
additional information, to the server via the communication
interface 2810. When the processor 2820 obtains image data and AI
data that correspond to the request, the processor 2820 may
determine whether to perform AI upscaling on the received image
data, based on the AI data. Based on a result of determining
whether to perform AI upscaling, the processor 2820 may perform AI
upscaling on the received image data through the DNN for upscaling
trained jointly with the DNN for downscaling of the server.
The processor 2820 may confirm a state of a network, based on a BER
or a timestamp of the image data received from the server via the
communication interface 2810. Based on the confirmed state of the
network, the processor 2820 may transmit a request for image data
of a different quality from among the plurality of items of image
data, based on the additional information, to the server via the
communication interface 2810. The processor 2820 may obtain the
image data and AI data corresponding to the request.
When the terminal 2800 includes a plurality of processors, each of
the processors may perform at least some of operations of the
processor 2820. For example, the CPU may confirm the state of the
network and may request image data corresponding thereto. The NPU
may perform AI upscaling on AI-encoded image data, and the GPU may
perform a process other than a process performed by the NPU, the
processes being included in the AI decoding process described with
reference to FIG. 2, or may support the NPU in performing the
process so as to accelerate the process performed by the NPU.
However, this is a an example, and operations to be performed by
the processors are not limited to the aforementioned examples.
The memory 2830 according to embodiments of the disclosure may
store various data, programs, or applications for driving and
controlling the terminal 2800. Each of the one or more programs
stored in the memory 2830 may include one or more instructions.
Each program (one or more instructions) or each application, which
is stored in the memory 2830, may be executed by the processor
2820.
Elements in a block diagram may be combined, an element may be
added thereto, or at least one of the elements may be omitted
according to actual specifications of an apparatus. That is, at
least two elements may be combined to one element, or one element
may be divided into two elements when necessary. Also, functions
performed by each element are for describing the embodiments of the
disclosure, and detailed operations or devices do not limit the
scope of the disclosure.
The aforementioned embodiments of the disclosure may be written as
computer-executable programs that may be stored in a medium.
The medium may continuously store the computer-executable programs,
or may temporarily store the computer-executable programs for
execution or downloading. Also, the medium may be any one of
various recording media or storage media in which a single piece or
plurality of pieces of hardware are combined, and the medium is not
limited to a medium directly connected to a computer system, but
may be distributed over a network. Examples of the medium include
magnetic media, such as a hard disk, a floppy disk, and a magnetic
tape, optical recording media, such as CD-ROM and DVD,
magneto-optical media such as a floptical disk, and ROM, RAM, and a
flash memory, which are configured to store program instructions.
Other examples of the medium include recording media and storage
media managed by application stores distributing applications or by
websites, servers, and the like supplying or distributing other
various types of software.
A model related to the DNN described above may be implemented as a
software module. When the DNN model is implemented as a software
module (for example, a program module including instructions), the
DNN model may be stored in a computer-readable recording
medium.
Also, the DNN model may be a part of at least one of the image
decoding apparatus, the image encoding apparatus, the server, or
the terminal described above by being integrated as a hardware
chip. For example, the DNN model may be manufactured as an
exclusive hardware chip for AI, or may be manufactured as a part of
an existing general-purpose processor (for example, CPU or AP) or a
graphic-exclusive processor (for example GPU).
Also, the DNN model may be provided as downloadable software. A
computer program product may include a product (for example, a
downloadable application) as a software program electronically
distributed through a manufacturer or an electronic market. For
electronic distribution, at least a part of the software program
may be stored in a storage medium or may be temporarily generated.
In this case, the storage medium may be a server of the
manufacturer or electronic market, or a storage medium of a relay
server.
The method and apparatus for streaming data according to
embodiments of the disclosure may transceive AI-encoded image data
by using a DNN, based on a state of a network, and thus may
constantly maintain QoS of reproduction of image data in a state of
the network which is changeable.
The effects that may be achieved by the method and apparatus for
streaming data according to embodiments of the disclosure are not
limited to the aforementioned features, and other unstated effects
will be clearly understood by one of ordinary skill in the art in
view of descriptions below.
While one or more embodiments of the disclosure have been described
with reference to the figures, it will be understood by one of
ordinary skill in the art that various changes in form and details
may be made therein without departing from the spirit and scope as
defined by the following claims.
* * * * *
References